10 papers · 1 filter
ProbePO: Sparse Probes, Redistributed Credit for Critic-Free LLM Reinforcement Learning
Diyi Hu, Ruizhong Qiu, Lin Qiu +2
Reinforcement learning from verifiable rewards assigns one terminal reward to an entire reasoning trajectory, obscuring how individual reasoning steps contributed to the outcome. G…
Residual Visual Credit Optimization: Conserved Evidence Routing for Multimodal Reinforcement Learning
Lin Qiu, Yao Liu, Diyi Hu +6
Reinforcement learning with verifiable rewards scales multimodal reasoning, but an outcome reward says how much a trajectory is worth, not how that value should be spread over the…
Self-Evolving Memory for Generative Recommendation
Xinyu Lin, Zhuosong Jiang, Zixiao Suo +12
Generative recommendation has emerged as a promising end-to-end paradigm for personalized recommendation. However, user preferences continuously evolve over time, making self-evolv…
Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall
Jacqueline He, Howard Yen, Shuyue Stella Li +9
Logit-based knowledge distillation (KD) is used to train smaller language models (LMs) via supervision from stronger teachers, but whether its benefits are consistent across traini…
EvoHarness-RL: Learning Runtime Harness Coordination for Self-Evolving Agents
Xuying Ning, Dongqi Fu, Tianxin Wei +17
Long-horizon LLM agents increasingly rely on external execution support to maintain state, track progress, recover from failures, and reuse experience across extended interactions.…
DAG-MoE: From Simple Mixture to Structural Aggregation in Mixture-of-Experts
Jiarui Feng, Hanqing Zeng, Karish Grover +11
Mixture-of-Experts (MoE) models have become a leading approach for decoupling parameter count from computational cost in large language models, yet effectively scaling MoE performa…
DUEL: Adversarial Self-Play for Multimodal Reasoning
Lin Qiu, Hanqing Zeng, Yao Liu +3
Reinforcement learning (RL) has emerged as an effective paradigm for improving the reasoning capability of vision-language models (VLMs). However, RL-based optimization typically d…
TARo: Token-level Adaptive Routing for LLM Test-time Alignment
Arushi Rai, Qiang Zhang, Hanqing Zeng +5
Large language models (LLMs) exhibit strong reasoning capabilities but typically require expensive post-training to reach high performance. Recent test-time alignment methods offer…
Text Has Curvature
Karish Grover, Hanqing Zeng, Yinglong Xia +2
Does natural language text have an intrinsic curvature? Language is increasingly modeled in curved geometries - hyperbolic spaces for hierarchy, mixed-curvature manifolds for compo…
Token-Level LLM Collaboration via FusionRoute
Nuoya Xiong, Yuhang Zhou, Hanqing Zeng +5
Large language models (LLMs) exhibit strengths across diverse domains. However, achieving strong performance across these domains with a single general-purpose model typically requ…