26 papers
PEAR: Adaptive Advantage Reshaping for Perception and Exploration in Large Vision-Language Models
Junxin Wang, Yingfeng Luo, Xiangchen Song +12
Effective multimodal reasoning requires a model to reliably utilize visual information while sufficiently exploring diverse reasoning paths. To achieve this goal, prior work combin…
Plan Canvas: Fixed Reasoning Regions for Continuous Language Flows
Miaohe Niu, Pengxiang Li, Jingbo Zhu +1
Continuous language flows generate text by denoising all positions of a target canvas together. The natural way to add reasoning to such a model is to write a trace ahead of the an…
Can Terminal Agents Trust Their Own Verification? Diagnosing and Improving Self-Verification
Yingfeng Luo, Shaowei Wei, Daixin Wang +7
Terminal agents rely on self-verification to assess and correct their solutions as they solve tasks through interaction with command-line environments. Yet how trustworthy such sel…
Complementary Roles of Activation and Parametric Memory in Few-Shot Learning
Miaohe Niu, Runsong Zhao, Xinyu Liu +5
At test time, large language models (LLMs) can encode historical information in activation memory (i.e., KV caches) and parametric memory (i.e., updated parameters). While activati…
Reading Right, Answering Wrong: How Visual Configuration Changes Affect Evidence Use in VLMs
Dingyang Lin, Yingfeng Luo, Chenglong Wang +4
Vision-language models (VLMs) have achieved strong performance on tasks such as visual question answering, yet small image resizes can turn correct answers into errors. We investig…
BabelArena: A Large-Scale Multilingual Benchmark for LLM Agents
Peng Kuang, Yuchun Fan, Jiangnan Li +7
Large language model (LLM) agents increasingly execute multi-step workflows through tool use and interaction with users and environments. However, current agent evaluations are lar…
Learning from Environmental Feedback: Credit Assignment across Multiple Timescales for Agentic Reinforcement Learning
Yifu Huo, Shunjie Xing, Chenglong Wang +8
Agentic reinforcement learning (RL) often suffers from delayed and sparse rewards in real-world environments. A promising solution to this challenge is credit assignment, which aim…
FlowCTS: On-policy Continuous Trajectory Supervision of Flow Models
Kaiyang Ye, Yuan Ge, Junxiang Zhang +8
While on-policy distillation (OPD) effectively addresses sparse rewards and exposure bias in large language model post-training, its extension to flow models remains underexplored.…
ToFu: A White-Box, Token-Efficient Agent Harness for Researchers
Junhao Ruan, Yuan Ge, Bei Li +7
Agentic coding tools present new opportunities to transform research workflows. The performance of agent systems built depends on both large language models (LLMs) and the harness…
HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps
Xin Liu, Runsong Zhao, Xinyu Liu +8
Extended chain-of-thought (CoT) traces improve LLM reasoning but incur substantial computational and memory costs. While existing CoT compression methods mitigate this by condensin…
Teacher-Guided Policy Optimization for On-Policy Reasoning Distillation under Large Policy Divergence
Xinyu Liu, Kechen Jiao, Chunyang Xiao +10
On-policy distillation (OPD) has become a promising paradigm for reasoning-oriented post-training of large language models (LLMs), especially when combined with reinforcement learn…
LANG: Reinforcement Learning for Multilingual Reasoning with Language-Adaptive Hint Guidance
Yuchun Fan, Bei Li, Peiguang Li +9
Reinforcement learning has proven effective for enhancing multi-step reasoning in large language models (LLMs), yet its benefits have not fully translated to multilingual contexts.…
MTR-Suite: A Framework for Evaluating and Synthesizing Conversational Retrieval Benchmarks
Junhao Ruan, Abudukeyumu Abudula, Bei Li +8
Accurate evaluation of conversational retrieval is pivotal for advancing Retrieval-Augmented Generation (RAG) systems. However, existing conversational retrieval benchmarks suffer…
StateRAG: Typed State Contracts for Complex Retrieval-Augmented Generation
Miaohe Niu, Pengxiang Li, Jingbo Zhu +1
Complex retrieval-augmented generation requires evidence retrieval and control over what to retrieve next, which paths to explore, whether evidence is sufficient, and which interme…
RouteLMT: Learned Sample Routing for Hybrid LLM Translation Deployment
Yingfeng Luo, Hongyu Liu, Dingyang Lin +6
Large Language Models (LLMs) have achieved remarkable performance in Machine Translation (MT), but deploying them at scale remains prohibitively expensive. A widely adopted remedy…
BIASEDTALES-ML: A Multilingual Dataset for Analyzing Narrative Attribute Distributions in LLM-Generated Stories
Yuxuan Ouyang, yingfeng luo, JingBo Zhu +1
Large Language Models (LLMs) are increasingly used to generate narrative content, including children's stories, which play an important role in social and cultural learning. Despit…
SPS: Steering Probability Squeezing for Better Exploration in Reinforcement Learning for Large Language Models
Yifu Huo, Chenglong Wang, Ziming Zhu +9
Reinforcement learning (RL) has emerged as a promising paradigm for training reasoning-oriented models by leveraging rule-based reward signals. However, RL training typically tends…
MSRL: Scaling Generative Multimodal Reward Modeling via Multi-Stage Reinforcement Learning
Chenglong Wang, Yifu Huo, Yang Gan +8
Recent advances in multimodal reward modeling have been largely driven by a paradigm shift from discriminative to generative approaches. Building on this progress, recent studies h…
Offline Exploration-Aware Fine-Tuning for Long-Chain Mathematical Reasoning
Yongyu Mu, Jiali Zeng, Fandong Meng +2
Through encouraging self-exploration, reinforcement learning from verifiable rewards (RLVR) has significantly advanced the mathematical reasoning capabilities of large language mod…
When Scaling Fails: Mitigating Audio Perception Decay of LALMs via Multi-Step Perception-Aware Reasoning
Ruixiang Mao, Xiangnan Ma, Dan Chen +12
Test-Time Scaling has shown notable efficacy in addressing complex problems through scaling inference compute. However, within Large Audio-Language Models (LALMs), an unintuitive p…
APR: Penalizing Structural Redundancy in Large Reasoning Models via Anchor-based Process Rewards
Kaiyan Chang, Chenwei Zhu, Yingfeng Luo +7
Test-Time Scaling (TTS) has significantly enhanced the capabilities of Large Reasoning Models (LRMs) but introduces a critical side-effect known as Overthinking. We conduct a preli…
SERM: Self-Evolving Relevance Model with Agent-Driven Learning from Massive Query Streams
Chenglong Wang, Canjia Li, Xingzhao Zhu +9
Due to the dynamically evolving nature of real-world query streams, relevance models struggle to generalize to practical search scenarios. A sophisticated solution is self-evolutio…
Probing Preference Representations: A Multi-Dimensional Evaluation and Analysis Method for Reward Models
Chenglong Wang, Yifu Huo, Yang Gan +10
Previous methods evaluate reward models by testing them on a fixed pairwise ranking test set, but they typically do not provide performance information on each preference dimension…
MRO: Enhancing Reasoning in Diffusion Language Models via Multi-Reward Optimization
Chenglong Wang, Yang Gan, Hang Zhou +10
Recent advances in diffusion language models (DLMs) have presented a promising alternative to traditional autoregressive large language models (LLMs). However, DLMs still lag behin…