2 citations · 3 across the 11 of their papers we have counts for
9 papers · 1 filter
Training-Free On-Policy Distillation
Bangji Yang, Jiajun Fan, Zhaoyu Zhu +2
On-policy distillation consolidates specialist models into one student at the price of rollouts, teacher passes and gradient steps for every teacher. We show that the consolidation…
: Spectral Null-Space Swap Makes Reasoning Models Efficient
Hongbo Ma, Sansheng Cao, Jiajun Fan +2
LLMs trained with Chain-of-thought excel in reasoning capability, but often come with excessive token cost. We find that the core of reasoning capacity lies in the Thinking model's…
SpeechGym: An Audio-Native Gym for Training Voice Agents via Reinforcement Learning
Jiajun Fan, Jingyuan Li, Prashanth Gurunath Shivakumar +6
Voice agents must call tools and hold multi-turn dialogue entirely through speech, yet the dominant paradigm trains them in text. Existing frameworks either cascade TTS and ASR aro…
Can We Read the Mind of an Audio LLM? A Verbalizable, Multilingual Middle-Layer Workspace
Jiajun Fan, Jingyuan Li, Prashanth Gurunath Shivakumar +8
An audio language model is a black box in a specific way: we see what it says, never what it works out on the way there, and chain-of-thought monitoring helps only if the model wri…
Constraint-First Reasoning: A Training-Free Protocol for Exploiting Answer-Space Constraints in Mathematical Problem Solving
Hongbo Ma, Bangji Yang, Yunqian Selina Cheng +3
Large language models can derive a plausible mathematical object yet still violate explicit requirements--for example, by omitting a modular reduction, returning a non-integer, or…
The Many Faces of On-Policy Distillation: Pitfalls, Mechanisms, and Fixes
Siqi Zhu, Xuyan Ye, Hongyu Lu +2
On-policy distillation (OPD) and on-policy self-distillation (OPSD) have emerged as promising post-training methods for large language models, offering dense token-level supervisio…
Retrieval is Cheap, Show Me the Code: Executable Multi-Hop Reasoning for Retrieval-Augmented Generation
Jiashuo Sun, Jimeng Shi, Yixuan Xie +10
Retrieval-Augmented Generation (RAG) has become a standard approach for knowledge-intensive question answering, but existing systems remain brittle on multi-hop questions, where so…
Batched Contextual Reinforcement: A Task-Scaling Law for Efficient Reasoning
Bangji Yang, Hongbo Ma, Jiajun Fan +1
Large Language Models employing Chain-of-Thought reasoning achieve strong performance but suffer from excessive token consumption that inflates inference costs. Existing efficiency…
Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation
Jiashuo Sun, Pengcheng Jiang, Saizhuo Wang +13
Retrieval-Augmented Generation (RAG) systems remain brittle under realistic retrieval noise, even when the required evidence appears in the top-K results. A key reason is that retr…