9 papers
ReasonAlloc: Hierarchical Decoding-Time KV Cache Budget Allocation for Reasoning Models
Wenhao Liu, Hao Shi, Yunhe Li +7
Long chain-of-thought (CoT) trajectories in large language model (LLM) reasoning cause severe inference bottlenecks due to rapid key-value (KV) cache growth. Current decoding-time…
Unifying Value Alignment and Assignment in Cross-Domain Offline Reinforcement Learning with Heterogeneous Datasets
Zhongjian Qiao, Jiafei Lyu, Chenjia Bai +3
Cross-domain offline reinforcement learning (RL) aims to learn a policy in the target domain with a limited target domain dataset and a source domain dataset that exhibits a dynami…
CP-Agent: A Calibrated Risk-Controlled Agent for Feedback-Driven Competitive Programming
Peisong Wang, Bowen Liu, Zehua Li +4
Large language models still struggle with contest-level programming, while many agentic remedies rely on massive inference-time sampling or expensive multi-stage post-training. We…
Astro: Activation-guided Structured Regularization for Outlier-Robust LLM Post-Training Quantization
Xi Chen, Ming Li, Junxi Li +5
Weight-only post-training quantization (PTQ) is crucial for efficient Large Language Model (LLM) deployment but suffers from accuracy degradation caused by weight and activation ou…
CoTBox-TTT: Grounding Medical VQA with Visual Chain-of-Thought Boxes During Test-time Training
Jiahe Qian, Yuhao Shen, Zhangtianyi Chen +2
Medical visual question answering could support clinical decision making, yet current systems often fail under domain shift and produce answers that are weakly grounded in image ev…
RLVER: Reinforcement Learning with Verifiable Emotion Rewards for Empathetic Agents
Peisong Wang, Ruotian Ma, Bang Zhang +13
Large language models (LLMs) excel at logical and algorithmic reasoning, yet their emotional intelligence (EQ) still lags far behind their cognitive prowess. While reinforcement le…