Prépublication
· 2026 arXiv
EN
Training language agents is bottlenecked by the high cost of long-horizon rollouts, yet standard agentic reinforcement learning (RL) learns only from a terminal scalar reward, leaving the supervision in intermediate steps largely unused. Prior work exploits th...
Haofei Yu, Yixing Chen et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Scaling up transformers has produced models with broad capabilities, deployed across heterogeneous hardware and fluctuating workloads. Yet, model families offer only a few fixed sizes (e.g., S, B, L), forcing users to commit to a model based on hardware and la...
Walter Simoncini, Dheeraj Varghese et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
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 prompt, which is the expectation of the reward for its response minus its inference costs. First-price...
Shubhendu Trivedi, Jacob Eisenstein et al.
#cs.GT
Prépublication
· 2026 arXiv
EN
We introduce Continuum Attention (CA), a causal Transformer attention operator that replaces discrete key and value lookup with integration over a continuous, piecewise linear memory curve. Each query induces an exponential relevance density along the curve. I...
Afshin Khadangi
#cs.LG
Prépublication
· 2026 arXiv
EN
Computer-use agents have demonstrated strong capabilities in interpreting screens and executing keyboard and mouse actions. However, games, with their visually rich worlds and diverse mechanics, remain challenging and offer a demanding next step toward general...
Mingyu Ouyang, Yiran Wang et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Long contexts enlarge key--value (KV) caches, increasing memory use and attention costs during generation. Compressing these caches must preserve information for unseen questions. We introduce **Latent Compaction**, which learns a compact document-specific cac...
Yingfa Chen, Xingyu Shen et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
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 action or changing the notation, leave the underlying task unchanged yet make the agent worse at it...
Artin Tajdini, Manpreet Kaur et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
We introduce a novel method for verifying Lean formalizations that goes beyond LLM-judged faithfulness. Unlike most verification methods, which rely on LLMs or trained models, our method combines LLM-based semantic review with mechanical numerical and vacuity...
Anna Mei, Michael R. Douglas
#cs.LO
Prépublication
· 2026 arXiv
EN
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 correct answer from any prefix the student writes. We call this property the recoverability of the teache...
Kaixiang Zhao, Tianrun Yu et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
Brazilian Portuguese remains underserved by open-source language models, and the few that are safety-aligned rarely document what did not work or validate comparisons with statistical significance. We present Manacá-1B-Instruct-2, the second-generation post-tr...
Bruno Leonardo Santos Menezes, Carlos Leonardo Souza Cardoso et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Automated research with LLM agents has shown promising results in various domains. Through iterative refinement and scored verifiers, LLM agents are capable of discovering new ideas. There are different ways to scale automated research: let one agent iterate f...
Shangyin Tan, Ben Zhou et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
3D object composition is pivotal to digital content creation and robot simulation, while also serving as a critical testbed for physical intelligence in foundation models. Previous methods either leverage large language models to directly predict object coordi...
Weichen Lyu, Chong Xia et al.
#cs.AI