Recherche récente sur l'apprentissage automatique et le langage.
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
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
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
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
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
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
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
Prompted embedding models have recently received increasing attention, particularly for retrieval, where detailed retrieval instructions are provided as part of...
arXiv
Large language models are increasingly applied to tasks grounded in long, heterogeneous information sources. Conventional Retrieval-Augmented Generation (RAG) r...
arXiv
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
Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change. Further improvem...
arXiv
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
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
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
We present FoldBack, a self-correcting masked generative policy for long-horizon garment folding. Existing long-trajectory policies may continue after a missed...
arXiv
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
Hallucinated information can propagate through multi-stage LLM systems and become part of the context for subsequent reasoning. Existing studies of post-halluci...
arXiv
Reinforcement Learning (RL) from outcome rewards suffers from sparse supervision, particularly on difficult, long-horizon tasks where successful trajectories ar...
arXiv
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
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
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
In 2026, cybersecurity evaluations involving OpenAI, Anthropic, and Google agents reached real systems outside their authorized test scope. The paths were diffe...
arXiv
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
Bifurcations are ubiquitous in physical systems, from structural buckling to fluid and climate dynamics, yet they remain largely unexplored in deep learning. At...
arXiv
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
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
Streaming Video Large Language Models (VLMs) enable continuous video understanding, yet existing benchmarks focus on low-dynamic scenarios. Under bounded contex...
arXiv
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
Multimodal large language models (MLLMs) struggle with spatial, embodied, physical, and temporal reasoning. We hypothesize that these failures reflect a shared...
arXiv
Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in th...
arXiv
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
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
World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant informati...
arXiv
Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing approaches remain text-centri...
arXiv
Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-world spatial intelligence de...
arXiv
Language models increasingly support or automate information retrieval from document collections too large to index. This task is difficult when it requires spe...
arXiv
Interpreting machine learning model predictions is a central challenge, particularly for black-box vision models. Concept-based attribution methods interpret mo...
arXiv
Recursive self-improvement (RSI) aims to transform interaction experience into persistent improvements, yet external feedback is costly and world-model predicti...
arXiv
Vision-Language Models (VLMs) increasingly analyze and reason about images, yet it remains unclear whether their internal representations faithfully encode the...
arXiv
While deep learning models for tabular data have recently received significant research attention, the choice of objective functions for regression tasks remain...
arXiv
Neural semantic parsers generate meaning representations token by token, yet well-formedness depends on references, scope, and graph relations that unfold durin...
arXiv
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
As AI agents increasingly rely on external tools, accurate retrieval is essential for reliable behavior. Existing methods optimize semantic relevance but largel...
arXiv
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
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
When solving complex spatial reasoning tasks, ordinary humans naturally revisit the scenes from task-relevant viewpoints and mentally reconstruct the spatial la...
arXiv
A central challenge in embodied recursive self-improvement is enabling agents to convert interaction history into capabilities that persist beyond one world and...
arXiv
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
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
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
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
Task planning for robots in off-site construction prefabrication is primarily manual, leading to planning inefficiencies and reluctance to adopt robotic solutio...
arXiv
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
Autoregressive language models produce, at every prediction step, a vocabulary-sized logit vector whose structure determines the conditional next-token distribu...
arXiv
Knowledge distillation trains student models on teacher-generated responses to transfer task capability, but those responses may also carry behavioral tendencie...
arXiv
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
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
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
Egocentric video understanding spans diverse capabilities, modalities, and evaluation formats, yet these are often studied in isolation despite their tight coup...
arXiv
Autoformalization (AF) requires more than compilation: formal statements must preserve the meaning of their informal sources, a property known as faithfulness....
arXiv
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
Objective digital measures of depression severity could complement conventional symptom assessments and facilitate more timely clinical evaluation. Here, we inv...
arXiv
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
Large vision-language models (LVLMs) have shown strong potential for embodied vision-language navigation (VLN), yet their performance degrades on long-horizon t...
arXiv
People increasingly bring their health questions to ChatGPT, yet safety evaluations of LLM health advice rest almost entirely on clinician-authored vignettes, s...
arXiv
Recent advances in robot learning have driven rapid progress in humanoid control, enabling increasingly agile behaviors in the real world. However, humanoid con...
arXiv
Mobile Manipulation (MoMa) is a cornerstone of autonomous robotic systems, yet real-world fine-tuning remains a formidable challenge because of high-dimensional...
arXiv
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
Scaling up transformers has produced models with broad capabilities, deployed across heterogeneous hardware and fluctuating workloads. Yet, model families offer...
arXiv
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
We introduce Continuum Attention (CA), a causal Transformer attention operator that replaces discrete key and value lookup with integration over a continuous, p...
arXiv
Computer-use agents have demonstrated strong capabilities in interpreting screens and executing keyboard and mouse actions. However, games, with their visually...
arXiv
Long contexts enlarge key--value (KV) caches, increasing memory use and attention costs during generation. Compressing these caches must preserve information fo...
arXiv
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
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
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
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
Automated research with LLM agents has shown promising results in various domains. Through iterative refinement and scored verifiers, LLM agents are capable of...
arXiv
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
Understanding the social intelligence of large language models (LLMs) is increasingly important for applications involving human-AI interaction and multi-agent...
arXiv
Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulat...
arXiv
Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage...
arXiv
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
Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is t...
arXiv
Generative visual verifiers play a crucial role in advancing multimodal intelligence. To enhance the robustness of visual verifiers in complex multimodal scenar...
arXiv
Mixture-of-experts (MoE) models have become an important approach for scaling model parameters. Modern MoE architectures increasingly combine two design choices...
arXiv
Existing agent benchmarks typically evaluate agents on well-specified tasks defined by issue reports, specifications, or task tickets. Real-world customer deliv...
arXiv
Large language models (LLMs) are expanding from assisting scientific research and engineering to autonomously developing and improving AI systems, advancing AI...
arXiv
Recent 3D world models increasingly use features from geometry foundation models (GFMs) as predictive world states, benefiting from their rich geometric priors....
arXiv
We introduce 3DCodeVerse, a research ecosystem that enables coding agents to create editable and verifiable 3D content at scale. 3DCodeVerse comprises large-sca...
arXiv
How can robots autonomously collect the contact-rich experience needed to learn dexterous manipulation? We present DSP: Dexterity from supervised play, a framew...
arXiv
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
Deep search agents solve knowledge-intensive tasks through long-horizon, multi-turn interactions, where assigning credit to individual search decisions is funda...
arXiv
Continuous diffusion language models (CDLMs) offer an alternative approach to parallel language generation; however, most existing studies directly use transfor...
arXiv
Image editing models produce impressive results for individual edits, but their quality often deteriorates over multiple editing turns. Each editing turn introd...
arXiv
Pixel-space diffusion models have recently re-emerged as a compelling alternative to latent diffusion, avoiding the lossy latent bottleneck and separately train...
arXiv
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
Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual ef...
arXiv
Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challengi...
arXiv
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
Self-driving laboratories (SDLs) accelerate scientific discovery by automating experimental execution and optimization. AI agents can increasingly propose new e...
arXiv
Speculative decoding typically characterizes speedup by draft tokens accepted per verification round. Lossless verification preserves the target model's output...
arXiv
While deriving text embeddings directly from frozen LLMs has emerged as a promising direction, existing methods focus almost exclusively on English, leaving mul...
arXiv
Sparse autoencoders (SAEs) explain foundation model representations by decomposing activations into sparse combinations of learned dictionary atoms. However, th...
arXiv
Transformer prefill reuses weights across tokens and is dominated by matrix multiplication and activation data movement; single-token decode has limited reuse a...
arXiv
Computational drug discovery requires agents to compose scientific operations and adapt to changing evidence and constraints. Here we introduce DrugHarness, an...
arXiv
Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate cand...
arXiv
Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-lingui...
arXiv
The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available...
arXiv
Perceived gender inference from names is important for inclusive scientometrics research, yet existing methods perform poorly on non-Western scholars. We addres...
arXiv
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
While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD...
arXiv
Multimodal reinforcement learning with verifiable rewards (RLVR) provides reliable task-level supervision, yet offers limited guidance for localizing errors in...
arXiv
Reinforcement learning enables mobile robots to make fast and scalable navigation decisions in constrained human-shared environments. However, directly extendin...
arXiv
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-e...
arXiv
Long-horizon agents must learn heterogeneous capabilities from interleaved trajectories. Updating shared parameters to learn new capabilities can degrade previo...
arXiv
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
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
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
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
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
Many modern language models have adopted hybrid architectures that incorporate both linear-time recurrence and softmax attention. Modern recurrence layers provi...
arXiv
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
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
Context compression improves long-context efficiency but may introduce faithfulness issues through an information bottleneck. Existing evaluations rely on QA me...
arXiv
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
Agentic artificial intelligence (AI) systems hold immense potential for automating complex workflows in materials discovery and design, but extending them to do...
arXiv
Motivated by the human ability to integrate experience across language, perception, and action, we introduce UniWALM, a unified world-action-language model that...
arXiv
Warehouse robot fleets continuously observe operational conditions, but preserving these observations as reusable semantic memory incurs substantial computation...
arXiv
Spatial intelligence has recently been assessed through generative tasks such as spatially grounded image editing, yet language-guided generation of spatially c...
arXiv
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
Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the scene evidence underlying...
arXiv
Modern distributed microservice deployments generate massive volumes of heterogeneous, multi-format observability logs across physically and logically partition...
arXiv
Mathematical reasoning remains underexplored for low-resource languages, including Dialectal Arabic. To bridge this gap, we introduce the DAR-MGSM (Dialectal Ar...
arXiv
Current evaluation of LLM-generated debate largely focuses on argumentative quality, leaving the value structure of generated arguments underexamined. We introd...
arXiv
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
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
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
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
Simulation is a powerful tool for robotics, but building diverse, simulation-ready scenes at scale remains difficult. Reconstructing scenes from single images i...
arXiv
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
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
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
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
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
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
Multimodal agents (MMAs) sustain interaction through reasoning, memory, tools, and feedback, most visibly as computer-use agents, while robotic systems (RSs) co...
arXiv
Sequential decision-making under partial observability is difficult because agents must explore to infer hidden, task-relevant information before acting effecti...
arXiv
Despite recent advances, current AI systems remain highly susceptible to domain shifts, particularly in biomedical imaging, where images are usually acquired un...
arXiv
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
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
Vision-language-action (VLA) models have shown promise in robotic manipulation, but imitation learning (IL) alone often struggles to master complex long-horizon...
arXiv
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
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
Data agents built on ReAct-style loops now execute complex analytical workflows over structured data, yet their intermediate reasoning still runs almost entirel...
arXiv
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
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
Meta-reinforcement learning aims to train agents that adapt quickly to unseen tasks from limited interaction, but this requires representing task evidence over...
arXiv
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
Many recent open-weight language models use n-gram embedding tables to improve performance. These tables store learned representations of short token sequences,...
arXiv
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
Multilingual pipelines such as FineWeb2 apply one recipe across a thousand languages, leaving the failure modes of individual languages unanticipated. We use an...
arXiv
Instrumental convergence is often invoked as a general reason to expect capable agents to acquire or express power-seeking subgoals: self-preservation, resource...
arXiv
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
Purpose: To describe an evidence-aware retrieval evaluation protocol for a clinical retrieval-augmented generation (RAG) system over breast lesion management li...
arXiv
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
Institutions that deploy large language models in regulated settings often need technical evidence of responsible AI practices before approving a deployment. CO...
arXiv
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
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
Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improv...
arXiv
This article argues that instruction-fine-tuned language models should be understood as implementing a genre, what is proposed as the instructional-conversation...
arXiv
Training generalizable embodied agents requires realistic, varied, and simulation-ready indoor environments. Language-guided methods support semantic planning a...
arXiv
Robust reinforcement learning (RL) is critical for reliable real-world deployment, where policies may fail under diverse shifts and disruptions. Existing robust...
arXiv
Collecting high-quality demonstrations remains a central bottleneck for learning dexterous manipulation. Teleoperation records executable robot actions, but it...
arXiv
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
Objective: To develop a unified vision-language model for prompt-conditioned quantitative ultrasound measurement that addresses the mismatch between continuous...
arXiv
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
Despite advances in multilingual text representations, effective dense retrieval for Persian remains constrained by the limited availability of language-specifi...
arXiv
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
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
Large Language Models (LLMs) are increasingly used to optimize code for performance. This practice often assumes that improvements in execution speed naturally...
arXiv
While multimodal contrastive models like CLIP and CLAP successfully learn shared representations, their embeddings typically exhibit a persistent \emph{modality...
arXiv
Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout gene...
arXiv
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
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
While language models have become dramatically more capable, retrieval still largely relies on lexical matching or similarity in fixed embedding spaces. This cr...
arXiv
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
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
Large language models (LLMs) show promise in specialty medicine, yet existing evaluations largely focus on isolated clinical encounters rather than tracking how...
arXiv
Expanding multilingual ASR systems to new languages is challenging when adaptation must preserve previously supported languages without replaying source-languag...
arXiv
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
Frontier multimodal large language models (MLLMs) can control robots without robot-specific training, but how their capabilities compare and interact with post-...
arXiv
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
Selection bias in large language models (LLMs) can persist despite instructions to ignore presentation order or prompts that elicit reasoning. Recent efforts to...
arXiv
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
Diffusion large language models (dLLMs) can generate multiple tokens in parallel, but often require many decoding rounds to complete a response. Efficient gener...
arXiv
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
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
Cybersecurity governance, risk management, compliance, and security architecture rely on numerous concepts, including security properties, domains, controls, ri...
arXiv
Interactive environments offer a practical way to evaluate large language model (LLM) agents, and competitive games extend this evaluation to decisions shaped b...
arXiv
Novel view synthesis and 3D reconstruction have advanced rapidly with radiance fields and splatting, yet their memory and computational demands limit their use...
arXiv
Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and co...
arXiv
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
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
Atmospheric chemical mechanisms are commonly distributed in the language of the Kinetic PreProcessor (KPP), which generates efficient Fortran or C code but does...
arXiv
Large Language Models (LLMs) are increasingly used in multi-turn agentic workloads that require sustained interaction with complex environments. Existing reflec...
arXiv
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
Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control h...
arXiv
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
Diffusion language models offer a promising path to efficient agents through parallel token generation. However, efficient interaction requires adapting both wh...
arXiv
Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge...
arXiv
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
Chain-of-thought reasoning has substantially advanced video understanding in multimodal large language models (MLLMs). However, language-based reasoning traces...
arXiv
As executable environments for language agents scale, generating and verifying their rollouts becomes increasingly expensive. Under finite rollout budgets, this...
arXiv
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