Prépublication
· 2026
As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left s...
arXiv
Prépublication
· 2026
Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discove...
arXiv
Prépublication
· 2026
Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs)...
arXiv
Prépublication
· 2026
Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a mo...
arXiv
Prépublication
· 2026
Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models t...
arXiv
Prépublication
· 2026
Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a principled way to es...
arXiv
Prépublication
· 2026
Language-model agents are increasingly asked to carry out open-ended scientific research, yet their results are usually graded against a known answer, a rubric,...
arXiv
Prépublication
· 2026
Prompted embedding models have recently received increasing attention, particularly for retrieval, where detailed retrieval instructions are provided as part of...
arXiv
Prépublication
· 2026
Large language models are increasingly applied to tasks grounded in long, heterogeneous information sources. Conventional Retrieval-Augmented Generation (RAG) r...
arXiv
Prépublication
· 2026
Many of the questions now put to large language models have no correct answer to score against: what a policy is worth, which option a user should choose, how t...
arXiv
Prépublication
· 2026
Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change. Further improvem...
arXiv
Prépublication
· 2026
How can we predict which base checkpoint is worth an expensive round of agentic post-training? End-to-end pass@$K$ tests whether successful behavior already app...
arXiv
Prépublication
· 2026
Deployments of research agents are moving to populations of thousands that share one pool of compute, while most current systems organize one project at a time...
arXiv
Prépublication
· 2026
AI learning tools are rapidly entering classrooms, but evidence about whether they help students learn is mixed and rests mostly on test scores. Comparatively l...
arXiv
Prépublication
· 2026
We present FoldBack, a self-correcting masked generative policy for long-horizon garment folding. Existing long-trajectory policies may continue after a missed...
arXiv
Prépublication
· 2026
Multi-teacher on-policy distillation (MOPD) is used in two settings. In common-domain composition, several teachers score each student rollout from one prompt d...
arXiv
Prépublication
· 2026
Hallucinated information can propagate through multi-stage LLM systems and become part of the context for subsequent reasoning. Existing studies of post-halluci...
arXiv
Prépublication
· 2026
Reinforcement Learning (RL) from outcome rewards suffers from sparse supervision, particularly on difficult, long-horizon tasks where successful trajectories ar...
arXiv
Prépublication
· 2026
Much knowledge work produces new deliverables from files a workspace already holds, and LLM agents are beginning to take such work over. Through direct corpus i...
arXiv
Prépublication
· 2026
Flow policies capture rich and diverse action distributions, and fine-tuning them with off-policy RL to improve beyond the demonstrations has drawn growing inte...
arXiv
Prépublication
· 2026
METR's 50\% time horizon measures the human completion time of software tasks that an AI solves with 50\% probability, allowing AI capabilities to be expressed...
arXiv
Prépublication
· 2026
In 2026, cybersecurity evaluations involving OpenAI, Anthropic, and Google agents reached real systems outside their authorized test scope. The paths were diffe...
arXiv
Prépublication
· 2026
We propose BrickBench, a benchmark for agentic text-conditioned LEGO-set design. Given a prompt, an agent is tasked with producing an assembly that not only sat...
arXiv
Prépublication
· 2026
Bifurcations are ubiquitous in physical systems, from structural buckling to fluid and climate dynamics, yet they remain largely unexplored in deep learning. At...
arXiv
Prépublication
· 2026
Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white-box deception detection v...
arXiv
Prépublication
· 2026
AI agents can now conduct real-world cyberattacks, scale up capabilities with the number of agents, and collectively pursue misaligned goals to obtain rewards....
arXiv
Prépublication
· 2026
Streaming Video Large Language Models (VLMs) enable continuous video understanding, yet existing benchmarks focus on low-dynamic scenarios. Under bounded contex...
arXiv
Prépublication
· 2026
A generalist robot should not only perform diverse tasks but also improve through experience, turning what it learns during execution into capabilities that lat...
arXiv
Prépublication
· 2026
Multimodal large language models (MLLMs) struggle with spatial, embodied, physical, and temporal reasoning. We hypothesize that these failures reflect a shared...
arXiv
Prépublication
· 2026
Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in th...
arXiv
Prépublication
· 2026
LLM developers post-train their models to exhibit prosocial values and behavioral traits, which are enumerated in an alignment target. However, while recent pos...
arXiv
Prépublication
· 2026
Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attributes, so models with sim...
arXiv
Prépublication
· 2026
World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant informati...
arXiv
Prépublication
· 2026
Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing approaches remain text-centri...
arXiv
Prépublication
· 2026
Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-world spatial intelligence de...
arXiv
Prépublication
· 2026
Language models increasingly support or automate information retrieval from document collections too large to index. This task is difficult when it requires spe...
arXiv
Prépublication
· 2026
Interpreting machine learning model predictions is a central challenge, particularly for black-box vision models. Concept-based attribution methods interpret mo...
arXiv
Prépublication
· 2026
Recursive self-improvement (RSI) aims to transform interaction experience into persistent improvements, yet external feedback is costly and world-model predicti...
arXiv
Prépublication
· 2026
Vision-Language Models (VLMs) increasingly analyze and reason about images, yet it remains unclear whether their internal representations faithfully encode the...
arXiv
Prépublication
· 2026
While deep learning models for tabular data have recently received significant research attention, the choice of objective functions for regression tasks remain...
arXiv
Prépublication
· 2026
Neural semantic parsers generate meaning representations token by token, yet well-formedness depends on references, scope, and graph relations that unfold durin...
arXiv
Prépublication
· 2026
Large language models reuse the same attention heads across tasks, but identifying these shared heads does not explain what they do differently in each task. A...
arXiv
Prépublication
· 2026
As AI agents increasingly rely on external tools, accurate retrieval is essential for reliable behavior. Existing methods optimize semantic relevance but largel...
arXiv
Prépublication
· 2026
Personal agents increasingly connect to services across the web and applications by using tools. As the effects of their actions accumulate and interact across...
arXiv
Prépublication
· 2026
Comics are one of the great visual records of the twentieth century: hundreds of thousands of pages in which a mass readership saw its leaders, its wars and its...
arXiv
Prépublication
· 2026
When solving complex spatial reasoning tasks, ordinary humans naturally revisit the scenes from task-relevant viewpoints and mentally reconstruct the spatial la...
arXiv
Prépublication
· 2026
A central challenge in embodied recursive self-improvement is enabling agents to convert interaction history into capabilities that persist beyond one world and...
arXiv
Prépublication
· 2026
Recent advances in automatic pronunciation assessment (APA) have driven a shift from scripted to free-form assessment. However, existing methods often rely on m...
arXiv
Prépublication
· 2026
Semantic query engines ask a large language model (LLM) to judge individual rows, and benchmarks compare these engines on the quality of a few executions at tem...
arXiv
Prépublication
· 2026
A useful language model should know when to answer and when to say "I don't know." Can we improve that decision by training the model itself or by training a se...
arXiv
Prépublication
· 2026
Multilingual dense retrievers place all languages in one shared space so that a question finds relevant documents, regardless of which language they are written...
arXiv
Prépublication
· 2026
Task planning for robots in off-site construction prefabrication is primarily manual, leading to planning inefficiencies and reluctance to adopt robotic solutio...
arXiv
Prépublication
· 2026
Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whethe...
arXiv
Prépublication
· 2026
Autoregressive language models produce, at every prediction step, a vocabulary-sized logit vector whose structure determines the conditional next-token distribu...
arXiv
Prépublication
· 2026
Knowledge distillation trains student models on teacher-generated responses to transfer task capability, but those responses may also carry behavioral tendencie...
arXiv
Prépublication
· 2026
Large language models (LLMs) increasingly act as personal assistants; however, little is known about how they capture national-level preferences on a global sca...
arXiv
Prépublication
· 2026
Buildings cause nearly 40\% of carbon emissions in the US. Of total building emissions, HVAC systems are responsible for approximately 40-60\% and account for 1...
arXiv
Prépublication
· 2026
Clinical trial protocols serve as the source of truth for study design, but are mostly written and used as unstructured PDF documents. Digitizing protocols turn...
arXiv
Prépublication
· 2026
Egocentric video understanding spans diverse capabilities, modalities, and evaluation formats, yet these are often studied in isolation despite their tight coup...
arXiv
Prépublication
· 2026
Autoformalization (AF) requires more than compilation: formal statements must preserve the meaning of their informal sources, a property known as faithfulness....
arXiv
Prépublication
· 2026
Tool-call judges evaluate argument values proposed by LLM agents. We find that they accept a correct value less often when their input holds no reference for it...
arXiv
Prépublication
· 2026
Objective digital measures of depression severity could complement conventional symptom assessments and facilitate more timely clinical evaluation. Here, we inv...
arXiv
Prépublication
· 2026
Terminal agents have shown strong abilities in complex coding tasks. However, they often struggle to recover from errors during long-running execution (e.g., co...
arXiv
Prépublication
· 2026
Large vision-language models (LVLMs) have shown strong potential for embodied vision-language navigation (VLN), yet their performance degrades on long-horizon t...
arXiv
Prépublication
· 2026
People increasingly bring their health questions to ChatGPT, yet safety evaluations of LLM health advice rest almost entirely on clinician-authored vignettes, s...
arXiv
Prépublication
· 2026
Recent advances in robot learning have driven rapid progress in humanoid control, enabling increasingly agile behaviors in the real world. However, humanoid con...
arXiv
Prépublication
· 2026
Mobile Manipulation (MoMa) is a cornerstone of autonomous robotic systems, yet real-world fine-tuning remains a formidable challenge because of high-dimensional...
arXiv
Prépublication
· 2026
Training language agents is bottlenecked by the high cost of long-horizon rollouts, yet standard agentic reinforcement learning (RL) learns only from a terminal...
arXiv
Prépublication
· 2026
Scaling up transformers has produced models with broad capabilities, deployed across heterogeneous hardware and fluctuating workloads. Yet, model families offer...
arXiv
Prépublication
· 2026
An agentic market is a decentralized serving-time market where AI agents bid to claim each user prompt. An agent's bid is based on its expected value for the pr...
arXiv
Prépublication
· 2026
We introduce Continuum Attention (CA), a causal Transformer attention operator that replaces discrete key and value lookup with integration over a continuous, p...
arXiv
Prépublication
· 2026
Computer-use agents have demonstrated strong capabilities in interpreting screens and executing keyboard and mouse actions. However, games, with their visually...
arXiv
Prépublication
· 2026
Long contexts enlarge key--value (KV) caches, increasing memory use and attention costs during generation. Compressing these caches must preserve information fo...
arXiv
Prépublication
· 2026
Post-training binds the competence of an LLM agent to the interface it was trained on. Semantics-preserving perturbations of that interface, such as renaming an...
arXiv
Prépublication
· 2026
We introduce a novel method for verifying Lean formalizations that goes beyond LLM-judged faithfulness. Unlike most verification methods, which rely on LLMs or...
arXiv
Prépublication
· 2026
On-policy distillation (OPD) trains a student model on its own rollouts with token-level supervision from a teacher, assuming that the teacher can reach a corre...
arXiv
Prépublication
· 2026
Brazilian Portuguese remains underserved by open-source language models, and the few that are safety-aligned rarely document what did not work or validate compa...
arXiv
Prépublication
· 2026
Automated research with LLM agents has shown promising results in various domains. Through iterative refinement and scored verifiers, LLM agents are capable of...
arXiv
Prépublication
· 2026
3D object composition is pivotal to digital content creation and robot simulation, while also serving as a critical testbed for physical intelligence in foundat...
arXiv
Prépublication
· 2026
Understanding the social intelligence of large language models (LLMs) is increasingly important for applications involving human-AI interaction and multi-agent...
arXiv
Prépublication
· 2026
Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulat...
arXiv
Prépublication
· 2026
Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage...
arXiv
Prépublication
· 2026
Predicting how the world responds to an action requires more than what can be seen: robots must account for environmental factors that are neither directly obse...
arXiv
Prépublication
· 2026
Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is t...
arXiv
Prépublication
· 2026
Generative visual verifiers play a crucial role in advancing multimodal intelligence. To enhance the robustness of visual verifiers in complex multimodal scenar...
arXiv
Prépublication
· 2026
Mixture-of-experts (MoE) models have become an important approach for scaling model parameters. Modern MoE architectures increasingly combine two design choices...
arXiv
Prépublication
· 2026
Existing agent benchmarks typically evaluate agents on well-specified tasks defined by issue reports, specifications, or task tickets. Real-world customer deliv...
arXiv
Prépublication
· 2026
Large language models (LLMs) are expanding from assisting scientific research and engineering to autonomously developing and improving AI systems, advancing AI...
arXiv
Prépublication
· 2026
Recent 3D world models increasingly use features from geometry foundation models (GFMs) as predictive world states, benefiting from their rich geometric priors....
arXiv
Prépublication
· 2026
We introduce 3DCodeVerse, a research ecosystem that enables coding agents to create editable and verifiable 3D content at scale. 3DCodeVerse comprises large-sca...
arXiv
Prépublication
· 2026
How can robots autonomously collect the contact-rich experience needed to learn dexterous manipulation? We present DSP: Dexterity from supervised play, a framew...
arXiv
Prépublication
· 2026
Large language models (LLMs) can often tell when they are being tested. This ability, known as evaluation awareness, can lead models to act more cautiously duri...
arXiv
Prépublication
· 2026
Deep search agents solve knowledge-intensive tasks through long-horizon, multi-turn interactions, where assigning credit to individual search decisions is funda...
arXiv
Prépublication
· 2026
Continuous diffusion language models (CDLMs) offer an alternative approach to parallel language generation; however, most existing studies directly use transfor...
arXiv
Prépublication
· 2026
Image editing models produce impressive results for individual edits, but their quality often deteriorates over multiple editing turns. Each editing turn introd...
arXiv
Prépublication
· 2026
Pixel-space diffusion models have recently re-emerged as a compelling alternative to latent diffusion, avoiding the lossy latent bottleneck and separately train...
arXiv
Prépublication
· 2026
Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it...
arXiv
Prépublication
· 2026
Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual ef...
arXiv
Prépublication
· 2026
Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challengi...
arXiv
Prépublication
· 2026
A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve...
arXiv
Prépublication
· 2026
Self-driving laboratories (SDLs) accelerate scientific discovery by automating experimental execution and optimization. AI agents can increasingly propose new e...
arXiv
Prépublication
· 2026
Speculative decoding typically characterizes speedup by draft tokens accepted per verification round. Lossless verification preserves the target model's output...
arXiv
Prépublication
· 2026
While deriving text embeddings directly from frozen LLMs has emerged as a promising direction, existing methods focus almost exclusively on English, leaving mul...
arXiv
Prépublication
· 2026
Sparse autoencoders (SAEs) explain foundation model representations by decomposing activations into sparse combinations of learned dictionary atoms. However, th...
arXiv
Prépublication
· 2026
Transformer prefill reuses weights across tokens and is dominated by matrix multiplication and activation data movement; single-token decode has limited reuse a...
arXiv
Prépublication
· 2026
Computational drug discovery requires agents to compose scientific operations and adapt to changing evidence and constraints. Here we introduce DrugHarness, an...
arXiv
Prépublication
· 2026
Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate cand...
arXiv
Prépublication
· 2026
Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-lingui...
arXiv
Prépublication
· 2026
The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available...
arXiv
Prépublication
· 2026
Perceived gender inference from names is important for inclusive scientometrics research, yet existing methods perform poorly on non-Western scholars. We addres...
arXiv
Prépublication
· 2026
Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through inter...
arXiv
Prépublication
· 2026
While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD...
arXiv
Prépublication
· 2026
Multimodal reinforcement learning with verifiable rewards (RLVR) provides reliable task-level supervision, yet offers limited guidance for localizing errors in...
arXiv
Prépublication
· 2026
Reinforcement learning enables mobile robots to make fast and scalable navigation decisions in constrained human-shared environments. However, directly extendin...
arXiv
Prépublication
· 2026
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-e...
arXiv
Prépublication
· 2026
Long-horizon agents must learn heterogeneous capabilities from interleaved trajectories. Updating shared parameters to learn new capabilities can degrade previo...
arXiv
Prépublication
· 2026
A seemingly reasonable push in Sokoban can make the puzzle unsolvable. To avoid such mistakes, an agent has to anticipate the consequences of its actions before...
arXiv
Prépublication
· 2026
The goal of a world model is to capture the dynamics of an environment. This has the great appeal that it allows agents to plan and simulate a course of actions...
arXiv
Prépublication
· 2026
Systems that combine several large language models (LLMs) often use one context for many questions. When the context is passed between models as text, the recei...
arXiv
Prépublication
· 2026
LLM agents that compute the right answer can still answer wrongly. Agent harnesses increasingly keep long tool outputs out of the context window, saving them to...
arXiv
Prépublication
· 2026
We propose KronLSTM, a linear attention architecture based on the Kronecker Delta Rule, a novel update rule for associative matrix memories. KronLSTM extends th...
arXiv
Prépublication
· 2026
Many modern language models have adopted hybrid architectures that incorporate both linear-time recurrence and softmax attention. Modern recurrence layers provi...
arXiv
Prépublication
· 2026
Frontier agents retrieve structural metadata well and compute geometry from coordinates less reliably. We ask how much of a benchmark score reflects that gap an...
arXiv
Prépublication
· 2026
Agent learning is shifting from learning from human data toward learning through interaction with the world. However, scaling such learning directly in real-wor...
arXiv
Prépublication
· 2026
Context compression improves long-context efficiency but may introduce faithfulness issues through an information bottleneck. Existing evaluations rely on QA me...
arXiv
Prépublication
· 2026
Video generation models can now depict remarkably realistic motion. Yet it is unclear how to use this motion beyond pixels: to animate a specific 3D asset and r...
arXiv
Prépublication
· 2026
Agentic artificial intelligence (AI) systems hold immense potential for automating complex workflows in materials discovery and design, but extending them to do...
arXiv
Prépublication
· 2026
Motivated by the human ability to integrate experience across language, perception, and action, we introduce UniWALM, a unified world-action-language model that...
arXiv
Prépublication
· 2026
Warehouse robot fleets continuously observe operational conditions, but preserving these observations as reusable semantic memory incurs substantial computation...
arXiv
Prépublication
· 2026
Spatial intelligence has recently been assessed through generative tasks such as spatially grounded image editing, yet language-guided generation of spatially c...
arXiv
Prépublication
· 2026
Multi-token prediction (MTP) can accelerate language-model decoding by drafting several tokens and verifying them in a single target-model pass. However, the be...
arXiv
Prépublication
· 2026
Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the scene evidence underlying...
arXiv
Prépublication
· 2026
Modern distributed microservice deployments generate massive volumes of heterogeneous, multi-format observability logs across physically and logically partition...
arXiv
Prépublication
· 2026
Mathematical reasoning remains underexplored for low-resource languages, including Dialectal Arabic. To bridge this gap, we introduce the DAR-MGSM (Dialectal Ar...
arXiv
Prépublication
· 2026
Current evaluation of LLM-generated debate largely focuses on argumentative quality, leaving the value structure of generated arguments underexamined. We introd...
arXiv
Prépublication
· 2026
TS2Vec is one of the strongest self-supervised methods for time series classification, introduced in 2022. It is designed to be trained anew on every dataset: s...
arXiv
Prépublication
· 2026
Computer-use agents are limited by the interfaces humans give them. An agent may type a spreadsheet cell by cell when a single line of code would do, or print a...
arXiv
Prépublication
· 2026
RTL repair benchmarks should measure whether a model can fix broken RTL given a specification. To this end, we construct and hash-bind RTLRepairDataset, 5,784 (...
arXiv
Prépublication
· 2026
Agentic on-policy distillation (OPD) trains a student agent on its own trajectories while a teacher scores every token. A natural explanation of where it goes w...
arXiv
Prépublication
· 2026
Simulation is a powerful tool for robotics, but building diverse, simulation-ready scenes at scale remains difficult. Reconstructing scenes from single images i...
arXiv
Prépublication
· 2026
Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning of large language models (LLMs), but it needs training data that are lo...
arXiv
Prépublication
· 2026
Existing sequential recommenders usually allocate a fixed computational budget to every item in a user's interaction history, regardless of the item's contribut...
arXiv
Prépublication
· 2026
Existing world models typically perceive a scene through current visual frame or a narrow history window. However, these observations may not reveal the physics...
arXiv
Prépublication
· 2026
Through a tangle of real-world interactions, from synthetic training pipelines to data shared on the open web, frontier LLMs absorb what their predecessors and...
arXiv
Prépublication
· 2026
Consensus is a fundamental problem in distributed computing and at the core of state machine replication (SMR). Despite decades of active human research, discov...
arXiv
Prépublication
· 2026
Quantitative research is an iterative loop: a researcher proposes a market hypothesis, implements it as an executable signal, tests it on historical data, and d...
arXiv
Prépublication
· 2026
Multimodal agents (MMAs) sustain interaction through reasoning, memory, tools, and feedback, most visibly as computer-use agents, while robotic systems (RSs) co...
arXiv
Prépublication
· 2026
Sequential decision-making under partial observability is difficult because agents must explore to infer hidden, task-relevant information before acting effecti...
arXiv
Prépublication
· 2026
Despite recent advances, current AI systems remain highly susceptible to domain shifts, particularly in biomedical imaging, where images are usually acquired un...
arXiv
Prépublication
· 2026
Many NP-complete (NPC) problems have a polynomial-time (P) twin that differs in a single structural detail, such as 3- versus 2-colouring a graph. A model that...
arXiv
Prépublication
· 2026
Reinforcement Learning with Human Feedback (RLHF) has become a central paradigm for aligning Large Language Models with human values. However, the reliance on s...
arXiv
Prépublication
· 2026
Vision-language-action (VLA) models have shown promise in robotic manipulation, but imitation learning (IL) alone often struggles to master complex long-horizon...
arXiv
Prépublication
· 2026
Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while generative models are semant...
arXiv
Prépublication
· 2026
Parallel AI agents can produce useful local edits yet degrade a shared solution when their edits are combined. We model this coordination problem as discrete bl...
arXiv
Prépublication
· 2026
Data agents built on ReAct-style loops now execute complex analytical workflows over structured data, yet their intermediate reasoning still runs almost entirel...
arXiv
Prépublication
· 2026
Test-time scaling has emerged as a key inference strategy for reasoning models, where a model generates a large number of candidate solutions in the hope of fin...
arXiv
Prépublication
· 2026
Wrong automatic routing of an IT support ticket can cost more than human triage. We evaluate cost-aware abstention as an offline, cost-conditional readiness ass...
arXiv
Prépublication
· 2026
Meta-reinforcement learning aims to train agents that adapt quickly to unseen tasks from limited interaction, but this requires representing task evidence over...
arXiv
Prépublication
· 2026
Tool-using language model agents rely on environmental facts to decide whether an action is permitted. When a tool response omits a safety-relevant fact, an age...
arXiv
Prépublication
· 2026
Many recent open-weight language models use n-gram embedding tables to improve performance. These tables store learned representations of short token sequences,...
arXiv
Prépublication
· 2026
Chain-of-thought (CoT) prompting elicits step-by-step rationales that appear to explain model predictions, but there is no guarantee these steps causally drive...
arXiv
Prépublication
· 2026
Multilingual pipelines such as FineWeb2 apply one recipe across a thousand languages, leaving the failure modes of individual languages unanticipated. We use an...
arXiv
Prépublication
· 2026
Instrumental convergence is often invoked as a general reason to expect capable agents to acquire or express power-seeking subgoals: self-preservation, resource...
arXiv
Prépublication
· 2026
While monolingual language model (LM) surprisal is widely established as a predictor of reading times, it cannot capture the non-native (L2) reading dynamics th...
arXiv
Prépublication
· 2026
Purpose: To describe an evidence-aware retrieval evaluation protocol for a clinical retrieval-augmented generation (RAG) system over breast lesion management li...
arXiv
Prépublication
· 2026
Control evaluations measure whether safeguards can prevent an AI model from causing undesired outcomes even when it is intentionally trying to do so. These eval...
arXiv
Prépublication
· 2026
Institutions that deploy large language models in regulated settings often need technical evidence of responsible AI practices before approving a deployment. CO...
arXiv
Prépublication
· 2026
Accuracy on medical examinations is common evidence offered for the clinical use of language models, yet a health system that intends to run an open model on it...
arXiv
Prépublication
· 2026
Research on sycophancy asks whether large language models (LLMs) bend towards what their users want to hear. We ask the reverse question: when an LLM is confide...
arXiv
Prépublication
· 2026
Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improv...
arXiv
Prépublication
· 2026
This article argues that instruction-fine-tuned language models should be understood as implementing a genre, what is proposed as the instructional-conversation...
arXiv
Prépublication
· 2026
Training generalizable embodied agents requires realistic, varied, and simulation-ready indoor environments. Language-guided methods support semantic planning a...
arXiv
Prépublication
· 2026
Robust reinforcement learning (RL) is critical for reliable real-world deployment, where policies may fail under diverse shifts and disruptions. Existing robust...
arXiv
Prépublication
· 2026
Collecting high-quality demonstrations remains a central bottleneck for learning dexterous manipulation. Teleoperation records executable robot actions, but it...
arXiv
Prépublication
· 2026
Personal email agents decide, for every message, what it is, whether it is urgent, whether to draft a reply, create a task, or ignore it. Their error rate is ra...
arXiv
Prépublication
· 2026
Objective: To develop a unified vision-language model for prompt-conditioned quantitative ultrasound measurement that addresses the mismatch between continuous...
arXiv
Prépublication
· 2026
Large language model (LLM) platforms introducing advertising must decide how much capability ad revenue should fund and what consumers should pay to avoid ads....
arXiv
Prépublication
· 2026
Despite advances in multilingual text representations, effective dense retrieval for Persian remains constrained by the limited availability of language-specifi...
arXiv
Prépublication
· 2026
Modern language models rely on subwording tokenization to define the discrete units of input processing, typically using methods like Byte Pair Encoding (BPE) t...
arXiv
Prépublication
· 2026
Statement autoformalization translates mathematics from natural language into a proof assistant such as Lean 4, and models now do it at scale. As provers get st...
arXiv
Prépublication
· 2026
Large Language Models (LLMs) are increasingly used to optimize code for performance. This practice often assumes that improvements in execution speed naturally...
arXiv
Prépublication
· 2026
While multimodal contrastive models like CLIP and CLAP successfully learn shared representations, their embeddings typically exhibit a persistent \emph{modality...
arXiv
Prépublication
· 2026
Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout gene...
arXiv
Prépublication
· 2026
Low-rank SVD-based compression offers a powerful strategy to reduce the computational costs of LLMs. However, existing methods face two key limitations: (i) glo...
arXiv
Prépublication
· 2026
In LLM-agent opinion simulations, each agent reports the probability that a claim is true after seeing its own previous probability and the probabilities report...
arXiv
Prépublication
· 2026
While language models have become dramatically more capable, retrieval still largely relies on lexical matching or similarity in fixed embedding spaces. This cr...
arXiv
Prépublication
· 2026
Evolution strategies (ES) update a model by sampling a population of perturbed copies, scoring each one, and moving in the reward-weighted direction of the pert...
arXiv
Prépublication
· 2026
Many everyday tasks require controlling contact force as well as motion. On a humanoid, this force must often be coordinated across both hands and the whole bod...
arXiv
Prépublication
· 2026
Large language models (LLMs) show promise in specialty medicine, yet existing evaluations largely focus on isolated clinical encounters rather than tracking how...
arXiv
Prépublication
· 2026
Expanding multilingual ASR systems to new languages is challenging when adaptation must preserve previously supported languages without replaying source-languag...
arXiv
Prépublication
· 2026
Human oversight is a pillar of AI safety. It is often exercised by a human in the loop (HITL) who approves critical actions taken by an AI agent. AI agents' inc...
arXiv
Prépublication
· 2026
Frontier multimodal large language models (MLLMs) can control robots without robot-specific training, but how their capabilities compare and interact with post-...
arXiv
Prépublication
· 2026
A vulnerability costs more to fix the later it is discovered. As coding agents generate code faster than developers can review it, software needs a first line o...
arXiv
Prépublication
· 2026
Selection bias in large language models (LLMs) can persist despite instructions to ignore presentation order or prompts that elicit reasoning. Recent efforts to...
arXiv
Prépublication
· 2026
A robot policy needs to understand what a task is asking for in a scene, and how to move the robot to accomplish that task. To achieve these two skills, we intr...
arXiv
Prépublication
· 2026
Diffusion large language models (dLLMs) can generate multiple tokens in parallel, but often require many decoding rounds to complete a response. Efficient gener...
arXiv
Prépublication
· 2026
Many academic disciplines (in particular, mathematics and the theoretical sciences) face difficult questions about how to adapt their practices to the rise of h...
arXiv
Prépublication
· 2026
Deep research, or any use of large language models (LLMs) to read a large document collection and produce insights, requires synthesizing information from diffe...
arXiv
Prépublication
· 2026
Cybersecurity governance, risk management, compliance, and security architecture rely on numerous concepts, including security properties, domains, controls, ri...
arXiv
Prépublication
· 2026
Interactive environments offer a practical way to evaluate large language model (LLM) agents, and competitive games extend this evaluation to decisions shaped b...
arXiv
Prépublication
· 2026
Novel view synthesis and 3D reconstruction have advanced rapidly with radiance fields and splatting, yet their memory and computational demands limit their use...
arXiv
Prépublication
· 2026
Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and co...
arXiv
Prépublication
· 2026
We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates....
arXiv
Prépublication
· 2026
Relations compose through the entity they share - Alice is Bob's mother and Bob is Carol's father, so Alice is Carol's grandmother - and the order matters, sinc...
arXiv
Prépublication
· 2026
Atmospheric chemical mechanisms are commonly distributed in the language of the Kinetic PreProcessor (KPP), which generates efficient Fortran or C code but does...
arXiv
Prépublication
· 2026
Large Language Models (LLMs) are increasingly used in multi-turn agentic workloads that require sustained interaction with complex environments. Existing reflec...
arXiv
Prépublication
· 2026
Scientific discovery often begins with an unexpected observation that prompts a new research question. In a world with altered physical laws, can large language...
arXiv
Prépublication
· 2026
Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control h...
arXiv
Prépublication
· 2026
Self-attention plays a key role in long-sequence modeling, but its quadratic time complexity fundamentally limits the scaling of Large Language Models. Recently...
arXiv
Prépublication
· 2026
Diffusion language models offer a promising path to efficient agents through parallel token generation. However, efficient interaction requires adapting both wh...
arXiv
Prépublication
· 2026
Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge...
arXiv
Prépublication
· 2026
When an LLM group's discussion is compressed into a short memory for a final decision maker, what should the memory keep? Two natural heuristics keep the rarest...
arXiv
Prépublication
· 2026
Chain-of-thought reasoning has substantially advanced video understanding in multimodal large language models (MLLMs). However, language-based reasoning traces...
arXiv
Prépublication
· 2026
As executable environments for language agents scale, generating and verifying their rollouts becomes increasingly expensive. Under finite rollout budgets, this...
arXiv
Prépublication
· 2026
For audio-language models, an answer change after a local sound-event edit can reflect a response to the event's class, sensitivity to waveform rewriting, or bo...
arXiv
Prépublication
· 2026
Simulated students offer a scalable way to test learning platforms and AI tutors before deployment with real students and to augment scarce data for studying le...
arXiv
Prépublication
· 2026
Memory systems for LLM agents are evaluated mostly on social dialogue, where facts accumulate and rarely change, so remembering reduces largely to finding the t...
arXiv
Prépublication
· 2026
Dynamic map (DM) is a core information infrastructure for vehicle-road-cloud (VRC) cooperative autonomous driving. It fuses perception data from multiple source...
arXiv
Prépublication
· 2026
Rubric optimization improves LLM grading by revising scoring instructions while keeping the grader's parameters fixed. Existing optimizers often judge edits by...
arXiv
Prépublication
· 2026
Surgical video understanding is essential for intraoperative assistance such as workflow monitoring, instrument preparation and procedural documentation. Delive...
arXiv
Prépublication
· 2026
Neural auction design has emerged as a powerful paradigm within automated mechanism design for modern Web platforms. Because these learned mechanisms directly d...
arXiv
Prépublication
· 2026
Large language model based agents increasingly rely on external tools to invoke services and interact with web resources. These tools can evolve through softwar...
arXiv
Prépublication
· 2026
Stress is a leading contributor to long-term physical and mental health decline. The automatic assessment of stress is a promising approach for continuous monit...
arXiv
Prépublication
· 2026
Frequent urban floodwater poses a serious threat to the safety and efficiency of intelligent transportation systems. Existing vision-based methods for floodwate...
arXiv
Prépublication
· 2026
On-policy distillation (OPD) can improve answer accuracy while weakening responsiveness to supporting evidence. We investigate this separation in scientific-doc...
arXiv
Prépublication
· 2026
Coding agents can now rewrite repository-scale codebases while preserving tested behavior, creating a new challenge for plagiarism detection, license auditing,...
arXiv
Prépublication
· 2026
Reverse and forward Kullback--Leibler (KL) divergences provide complementary supervision in on-policy distillation: reverse KL refines the student's active toke...
arXiv
Prépublication
· 2026
Traditional interactive narrative systems offer strong authorial control and structural coherence but often restrict player agency. In contrast, LLM-driven narr...
arXiv
Prépublication
· 2026
Existing attacks and defenses for coding agents act once content enters the model context: injected text persuades the model, and guards inspect what the model...
arXiv
Prépublication
· 2026
Networks that keep learning overwrite what they know. Prevailing remedies replay stored data, distil from old models or freeze the backbone. The brain suggests...
arXiv
Prépublication
· 2026
LLM-guided evolutionary search uses AI agents to improve programs for mathematical and computational problems by iteratively refining solutions and combining hi...
arXiv
Prépublication
· 2026
AI-Scientists arrived, but a world for them to grow in did not. Human scientists take theirs for granted $-$ centuries of open problems and shared ideas, waitin...
arXiv
Prépublication
· 2026
ML code should run where researchers choose, not where it was first written. Today, machine learning code mainly lives in two ecosystems, PyTorch on NVIDIA GPUs...
arXiv
Prépublication
· 2026
Interpretability, AI auditing and assessments of machine consciousness increasingly rest on mechanism-level quantities: functionals of a system's causal transit...
arXiv
Prépublication
· 2026
To interpret a direction in a language model's residual stream, we can look for text that strongly activates it. But when such text is found by scanning a corpu...
arXiv
Prépublication
· 2026
On-policy distillation (OPD) trains a student on its own rollouts under a frozen teacher's per-token supervision. Does this dense signal reduce the training FLO...
arXiv
Prépublication
· 2026
Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, released code, and supporting artifacts. F...
arXiv
Prépublication
· 2026
Typed-decision models score options that the caller supplies at request time, so their accuracy depends on label wording that the model's developer does not con...
arXiv
Prépublication
· 2026
Neural networks are likely to be among the most consequential artefacts in human history. As these systems become more capable and autonomous, their potential t...
arXiv
Prépublication
· 2026
LLM-based medical assistants that follow a patient across visits need a memory that keeps the patient's clinical state correct as it changes. We examine the err...
arXiv
Prépublication
· 2026
Chat remains the default surface for many large language model (LLM) tasks, even as agentic tools such as OpenCode excel at coding and automation. Yet the model...
arXiv
Prépublication
· 2026
Vision-Language-Action models (VLAs) enable general-purpose embodied intelligence, but their performance is brittle under distribution shift, where unanticipate...
arXiv
Prépublication
· 2026
Coding agents are among the most economically consequential deployments of LLMs to date, making reliable evaluation increasingly important. Yet even benchmarks...
arXiv
Prépublication
· 2026
The growing use of AI in mathematics raises a question: can we detect AI involvement in proofs, separately from AI assistance with writing? Historical arXiv pap...
arXiv
Prépublication
· 2026
Self-consistency (SC) is a widely used decoding strategy for improving chain-of-thought reasoning, but it requires sampling many complete reasoning trajectories...
arXiv
Prépublication
· 2026
We describe the system submitted by Telefónica Innovación Digital to the Speech-COSER Challenge 2026, which features spontaneous speech from elderly speakers re...
arXiv
Prépublication
· 2026
Existing evaluations often test persona expression rather than whether persona information persists in extended decision-making. We introduce THEATRON, a contro...
arXiv
Prépublication
· 2026
Swarm Learning improves resilience and removes the need for a single trusted coordinator by distributing model aggregation and validation across participating n...
arXiv
Prépublication
· 2026
The small open embedder EmbeddingGemma-300M, used as released, finds the right passage first for only 51.7% to 60.8% of Belebele questions in Tajik, Kazakh, Uzb...
arXiv
Prépublication
· 2026
Generating accurate CAD models requires substantial engineering expertise and careful control of geometry, topology, and mechanical constraints. Text-to-CAD aim...
arXiv
Prépublication
· 2026
Large language models (LLMs) perform well on general coding tasks, but their capabilities in application-layer embedded software development remain underexplore...
arXiv
Prépublication
· 2026
Aligning two models without paired data requires a distance that both share, yet existing methods use Euclidean or spherical geometry. This is either an arbitra...
arXiv
Prépublication
· 2026
Temporally correlated noise can substantially improve the utility of differentially private optimization, yet candidate mechanisms are commonly designed or eval...
arXiv
Prépublication
· 2026
On-policy distillation (OPD) trains a student language model on its own responses, guided by a fixed teacher. Generating these responses is expensive, so system...
arXiv
Prépublication
· 2026
The order in which propagators are executed can substantially affect the performance of constraint programming solvers, yet propagator priorities are typically...
arXiv
Prépublication
· 2026
Evaluating general-purpose multimodal large language models (MLLMs) for surgical scene understanding remains limited due to scarce surgical data and fragmented...
arXiv
Prépublication
· 2026
A legal answer is only as good as the article it rests on. Vietnam's largest open legal corpus pairs 109,543 published questions and answers with 573,851 legal...
arXiv
Prépublication
· 2026
Humans engage in pragmatic reasoning when encountering mentions of highly predictable events. Such mentions may appear redundant, leading comprehenders to infer...
arXiv
Prépublication
· 2026
Dense retrievers are meant to select passages that match a query's semantic specification. We show they suffer from search-by-answer bias: they favor passages t...
arXiv
Prépublication
· 2026
Cross-lingual concept analysis requires distinguishing the information that a subspace retains from the classifiers and erasers that transfer through it. We stu...
arXiv
Prépublication
· 2026
Foundation models pretrained on large natural-image collections are increasingly adapted for medical image segmentation, in the expectation that they will repla...
arXiv
Prépublication
· 2026
DeepLearning (DL)-based Computer-Aided Diagnostic (CAD) systems have shown great promise in supporting clinicians, but their performance heavily depends on the...
arXiv
Prépublication
· 2026
Reinforcement learning with verifiable rewards (RLVR) produces rich training experience, yet outcome rewards reduce each attempt to a single number and leave mo...
arXiv
Prépublication
· 2026
LLM agents are increasingly used for open-ended, data-driven scientific discovery, yet most of them evaluate hypotheses pointwise, testing an empirical claim $h...
arXiv
Prépublication
· 2026
Deploying LLM-based analytics agents in enterprise settings requires evaluation frameworks that can reliably detect failures across complex, multi-tool workflow...
arXiv
Prépublication
· 2026
Business process automation runs the same Standard Operating Procedure (SOP) again and again on new inputs. Large Language Model (LLM) agents can follow an SOP...
arXiv
Prépublication
· 2026
Research ideation is iterative: researchers refine their ideas step by step based on feedback. Yet existing evaluations obscure this search process, making it d...
arXiv
Prépublication
· 2026
Neural networks are increasingly studied not merely as predictive systems, but as complex systems whose internal computations are themselves objects of scientif...
arXiv
Prépublication
· 2026
In this paper, we present 1020 Group's submission to the WMT26 shared task on automated quality evaluation. Our system, Pterodactyl, was engineered specifically...
arXiv
Prépublication
· 2026
Text-to-SQL systems over large pooled catalogs must first retrieve the tables a question needs. We compare against the published table recall of RASL, which use...
arXiv
Prépublication
· 2026
UTF-8's role as the de facto encoding standard in language models introduces cross-lingual inequities due to its variable-length design. However, the effect tha...
arXiv
Prépublication
· 2026
We describe our solution for the 2026 BEHAVIOR Challenge: 100 long-horizon household tasks in the OmniGibson simulator. We placed 9th on the final leaderboard,...
arXiv
Prépublication
· 2026
Branch-and-Bound (B&B) and $A^*$ are popular exact search methods originating from Operations Research and Artificial Intelligence, respectively. Although the t...
arXiv
Prépublication
· 2026
Time-Series Foundation Models (TSFMs) are increasingly used in sensor applications, where sampling frequency (SF) varies across sensors, devices, and operating...
arXiv
Prépublication
· 2026
Streaming speech recognizers output a word only after the speaker has finished it, typically 90-400 ms later. For a voice agent every millisecond counts: the so...
arXiv
Prépublication
· 2026
Agents that call tools and robots that execute tasks are challenged by two shortcomings: they repeat what failed before, and they learn from runs that they only...
arXiv
Prépublication
· 2026
A coding agent working within a software organization must establish what to build, collaborate with colleagues, and verify what it delivers. We introduce SWE-C...
arXiv
Prépublication
· 2026
Subword tokenizers assume that meaningful linguistic units can be recovered by splitting text at learned boundaries, an assumption better suited to concatenativ...
arXiv
Prépublication
· 2026
Generative AI increasingly competes with creators while using their content as training data. In response, policymakers have called for platforms to compensate...
arXiv
Prépublication
· 2026
When LLMs turn requirements into code, every underspecified constraint becomes behavior that the model must decide during generation. We test when specification...
arXiv
Prépublication
· 2026
Typed Decision Models such as TypeSafe's Jev, also called System One or System 1 models, answer typed questions about an input with a probability for each optio...
arXiv
Prépublication
· 2026
The CAPSARII project advances a cyber-secure, ultra-low-power wearable platform and Internet of Battlefield Things (IoBT) framework to support real-time monitor...
arXiv
Prépublication
· 2026
We present RoboKit, a high-performance, general-purpose toolkit for parallel differentiable optimization in robotics, built on three principles: (i) Performance...
arXiv
Prépublication
· 2026
AI-text detectors can be sensitive to changes in writing style. We examined this sensitivity by rewriting 991 AI-generated answers with Gemini, using prompts th...
arXiv
Prépublication
· 2026
LLM-as-a-judge guardrails are often controlled primarily through system prompts, yet standard prompt optimization methods target average performance rather than...
arXiv
Prépublication
· 2026
Directional ablation removes a hidden state's component along a "behaviour direction" while a language model runs. It is the standard way to switch a behaviour...
arXiv
Prépublication
· 2026
UI grounding asks a model to find the screen element that a user instruction refers to. Most current systems use a vision-language model (VLM) that reads a scre...
arXiv
Prépublication
· 2026
Memory systems allow AI agents to persist information across interactions, making them more personalized and more capable on long-horizon tasks. This persistenc...
arXiv
Prépublication
· 2026
Delegating resume screening to large language models (LLMs) requires organizations to establish whether they can distinguish candidates by relevant qualificatio...
arXiv
Prépublication
· 2026
Determining whether an embedding is sufficiently similar to any vector in a vector store without performing a nearest-neighbor search can substantially improve...
arXiv
Prépublication
· 2026
Collecting large-scale, diverse, and high-quality post-training datasets for autoregressive language models is challenging. Consequently, Supervised Fine-Tuning...
arXiv
Prépublication
· 2026
We present StringQFT as a textbook collection for string constructions of quantum field theory and as infrastructure for AI-assisted theoretical physics. The co...
arXiv
Prépublication
· 2026
LLM agents increasingly drive their own learning. Self-improvement methods delegate various components of model optimization to the models themselves. As models...
arXiv
Prépublication
· 2026
Methods that localise social bias in large language models to individual attention heads and remove it by ablation are validated by their effect on a fairness m...
arXiv
Prépublication
· 2026
New coding models advertise themselves on score improvements of coding benchmarks. Yet no study has tested if a model that scores higher on these benchmarks get...
arXiv
Prépublication
· 2026
Vision-language models (VLMs) can provide harmful assistance by interpreting visual events, even without an explicit request for the underlying harmful objectiv...
arXiv
Prépublication
· 2026
Small language models that run on a laptop are attractive for writing signal-processing code offline, but they make systematic errors on digital signal processi...
arXiv
Prépublication
· 2026
AI agents increasingly collaborate across organizational boundaries, invoking tools and interacting with agents operated by independent entities. Although exist...
arXiv
Prépublication
· 2026
Next-token prediction is relied upon to learn both predictive behavior and internal representations. We ask whether representation structure itself should be an...
arXiv
Prépublication
· 2026
Goal recognition is the task of inferring an agent's intended goal from its observed behavior. Most existing work focuses on high-level symbolic recognition, ab...
arXiv
Prépublication
· 2026
Automated segmentation is routinely used as the first step in quantitative image-based clinical workflows, where predicted masks are subsequently reviewed and c...
arXiv
Prépublication
· 2026
We investigate how language models process negation in natural language inference (NLI). We propose the \textit{Polarity Composition Hypothesis}, a high-level c...
arXiv
Prépublication
· 2026
Preference optimization (PO) is the standard way to reduce object hallucination in large vision-language models. We find that four public PO checkpoints of LLaV...
arXiv
Prépublication
· 2026
Empirical research is changing rapidly as AI coding agents are becoming increasingly proficient in every step of it, from writing code to manipulating data. We...
arXiv
Prépublication
· 2026
Networks that predict angles, gaze directions, and rotations do not output these quantities directly: they emit raw coordinates that a fixed decoder, such as no...
arXiv