16 papers
Better Decomposition, Free Aggregation: A Synthesizer-Folding Framework for Multilingual Multi-Hop Question Answering
Yilin Wang, Yuchun Fan, Weidong Bao +5
Multilingual retrieval-augmented generation (mRAG) equips large language models with access to globally distributed external knowledge for complex multilingual question answering.…
DF-ReAG: Dynamic Decomposition and Filtering for Multi-Hop Reasoning-Augmented Generation
Jiaoyang Li, Junhao Ruan, Shengwei Tang +4
Large language models (LLMs) often generate inaccurate answers due to their reliance on static internal knowledge. Retrieval-augmented generation (RAG) addresses this limitation by…
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.…
Detecting Is Not Resolving: The Monitoring Control Gap in Retrieval Augmented LLMs
Zhe Yu, Wenpeng Xing, Chen Ye +4
Retrieval-augmented LLMs are deployed for tasks where evidence quality determines action safety, yet evaluation protocols assume that single-turn robustness predicts robustness whe…
Composition Collapse: Stable Factual Knowledge Does Not Imply Compositional Reasoning
Zhe Yu, Wenpeng Xing, Yunzhao Wei +4
Post-training is routinely evaluated through aggregate benchmark scores that treat multi-hop reasoning as a single capability -- as if a model that answers more questions correctly…
The Attribution Blind Spot: Detecting When Language Models Rely on Memory Rather Than Retrieved Context
Zhe Yu, Wenpeng Xing, Yunzhao Wei +4
Retrieval-augmented generation promises to ground language model outputs in external evidence, yet the field has no reliable way to verify whether retrieved context actually govern…
Cordon-MAS: Defending RAG against Knowledge Poisoning via Information-Flow Control
Zhe Yu, Wenpeng Xing, Gaolei Li +4
Retrieval-augmented generation (RAG) increasingly underpins high-stakes applications, yet remains vulnerable to Confundo-style poisoning where adversarially optimized documents man…
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…
On the Emotion Understanding of Synthesized Speech
Yuan Ge, Haishu Zhao, Aokai Hao +10
Emotion is a core paralinguistic feature in voice interaction. It is widely believed that emotion understanding models learn fundamental representations that transfer to synthesize…
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…
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…
M-CIF: Multi-Scale Alignment For CIF-Based Non-Autoregressive ASR
Ruixiang Mao, Xiangnan Ma, Qing Yang +7
The Continuous Integrate-and-Fire (CIF) mechanism provides effective alignment for non-autoregressive (NAR) speech recognition. This mechanism creates a smooth and monotonic mappin…
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…
MTP-S2UT: Enhancing Speech-to-Speech Translation Quality with Multi-token Prediction
Jianjin Wang, Runsong Zhao, Xiaoqian Liu +6
Current direct speech-to-speech translation methods predominantly employ speech tokens as intermediate representations. However, a single speech token is not dense in semantics, so…
Autoencoding-Free Context Compression for LLMs via Contextual Semantic Anchors
Xin Liu, Runsong Zhao, Pengcheng Huang +7
Context compression is an advanced technique that accelerates large language model (LLM) inference by converting long inputs into compact representations. Existing methods primaril…
FLEXI: Benchmarking Full-duplex Human-LLM Speech Interaction
Yuan Ge, Saihan Chen, Jingqi Xiao +5
Full-Duplex Speech-to-Speech Large Language Models (LLMs) are foundational to natural human-computer interaction, enabling real-time spoken dialogue systems. However, benchmarking…