Policy optimisation, reward modelling and RLHF — how models are trained by feedback rather than by labels.
arXiv:2608.22167v1 Announce Type: new Abstract: Reinforcement learning (RL) has become an effective way to improve the tool-use ability of large language models (LLMs), but most existing RL framework...
arXiv:2608.21898v1 Announce Type: new Abstract: Web agents promise to automate complex digital workflows, but their training remains limited by synthetic environments that look plausible while hiding...
arXiv:2608.21372v1 Announce Type: new Abstract: This report summarizes the work conducted on the Game of Hidden Rules (GOHR), focusing on reinforcement learning agents trained to infer hidden rules f...
arXiv:2608.23329v1 Announce Type: cross Abstract: Open-world video understanding often requires a model to locate sparse visual evidence and acquire external knowledge that is absent from the video a...
DeepRefine is a reinforcement learning framework that improves the quality of pre‑constructed structured knowledge bases—such as knowledge graphs or LLM‑Wikis—by engaging in multi‑turn interactions with the base. It performs abductive diagnosis to locate defects, then applies targeted refinement actions to incrementally update the knowledge base. The system uses a Gain‑Beyond‑Draft reward to train its refinement policy end‑to‑end, achieving consistent downstream performance gains over strong baselines.
The paper presents a reward‑free continual learning framework for space robots that uses latent‑state world models to adapt to severe hardware degradation. By pre‑training a model‑based agent in diverse simulations, the world model learns to predict reward structure in latent space. During deployment, the observation encoder and reward predictor are frozen while only the transition dynamics are updated via unsupervised rollouts, allowing the policy to adapt using imagined trajectories without new rewards.
VisionCoach is an input‑adaptive reinforcement learning framework that enhances spatio‑temporal grounding in video reasoning by using visual prompting during training. The system selectively applies visual prompts to challenging inputs, amplifying question‑relevant evidence and suppressing distractors, and then internalizes these improvements through self‑distillation so that inference can be performed on raw videos without prompts. Experiments on multiple benchmarks (V‑STAR, VideoMME, World‑Sense, VideoMMMU, PerceptionTest, and Charades‑STA) show that VisionCoach achieves state‑of‑the‑art performance while maintaining a single efficient inference pathway.
The paper introduces MEMENTO, a memory‑enhanced neural solver that improves routing problem solutions by using online data from repeated attempts to adjust action distributions during inference. It targets NP‑hard routing tasks such as the Traveling Salesman and Capacitated Vehicle Routing problems, outperforming existing tree‑search and policy‑gradient fine‑tuning methods. MEMENTO demonstrates strong scalability and data efficiency, achieving state‑of‑the‑art results on 11 of 12 evaluated tasks and enabling zero‑shot integration with diversity‑based solvers.
arXiv:2608.23318v1 Announce Type: new Abstract: Hint-based reinforcement learning addresses reward sparsity in long-horizon agentic tasks by retaining a prefix of an expert trajectory before each rol...
The paper introduces Ask-Condition-Abstain Reinforcement Learning (ACA‑RL), a framework that trains reasoning models to handle queries missing a premise by either asking for it, conditioning on the unknown, or abstaining. ACA‑RL uses a reasoning‑graph‑guided pipeline to generate training instances with localized gap annotations and a structured reward over five observable response behaviors. The authors also present the Missing‑Premise Benchmark (MPB), a 274‑instance, human‑verified dataset covering mathematical, logical, and real‑world word problems, and show that ACA‑RL improves performance on MPB while maintaining competitive results on well‑posed tasks for Qwen3 and Llama models.
The paper introduces PM4Bench, a multimodal, multilingual, multi-task benchmark built on a strictly parallel 10‑language corpus, allowing fair cross‑lingual comparison of Large Vision‑Language Models (LVLMs). It also proposes a vision setting that embeds textual inputs directly into images to better mimic real deployment scenarios. Experiments show OCR performance drives cross‑lingual gaps, leading to an OCR‑centric GRPO training strategy that improves multilingual VQA and reduces disparities without costly supervision.
arXiv:2608.22364v1 Announce Type: new Abstract: World action models (WAMs) couple visual future prediction with robot action generation, but accelerated students can lose task capabilities during dis...
arXiv:2608.21501v1 Announce Type: new Abstract: Credit assignment in large-language-model reinforcement learning (LLM RL) can be separated into three objects: evidence about success, a transport oper...
arXiv:2608.21595v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) improves the reasoning ability of vision-language models (VLMs), and diversifying the rollouts wi...
The paper introduces Delta, a two‑phase framework for testing deep reinforcement learning agents. In the first phase, the agent under test is evaluated for catastrophic failures while collecting decision‑making data. The second phase trains a challenger agent from this data using offline RL; comparing the challenger’s rewards to the original agent reveals optimality bugs, and Delta successfully uncovered thousands of such issues across multiple environments.
arXiv:2608.21860v1 Announce Type: cross Abstract: Chain-of-Thought (CoT) reasoning has significantly enhanced the multi-step problem-solving capabilities of large language models (LLMs) by introducin...
arXiv:2606.04588v2 Announce Type: replace Abstract: Multimodal large language models have made rapid progress in video understanding, yet existing benchmarks largely rely on simple prompts and provid...
arXiv:2608.21386v1 Announce Type: cross Abstract: Given a task described by a few examples, how should a model be specialized to it? Four mechanisms are available -- zero-shot, in-context attention,...
arXiv:2608.21811v1 Announce Type: new Abstract: Reinforcement learning for vision-language math reasoning starves under sparse reward: on a pool of 20,830 visual-math problems where Qwen2-VL-2B answe...
arXiv:2608.21830v1 Announce Type: new Abstract: Graphical User Interface (GUI) agents powered by Multimodal Large Language Models (MLLMs) have shown strong potential for automating tasks across diver...