15 papers · 1 filter
Distillation as Probability Transport: Routed On-Policy Distillation
Tianle Xia, Lingxiang Hu, Yiding Sun +6
On-policy distillation (OPD) transfers teacher knowledge on student-generated trajectories, but efficient sampled objectives reduce the teacher distribution to scalar credit on ind…
Teaching a Moving Student: Rethinking the Curriculum of On-Policy Distillation
Lingxiang Hu, Tianle Xia, Ming Xu +6
In on-policy distillation (OPD), the student determines which states receive teacher supervision. As its policy evolves, earlier response prefixes become less likely even though te…
When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents
Linfang Shang, Ming Xu, Yiding Sun +4
Self-evolving agents accumulate capability by distilling reusable skills from their execution trajectories, but we find this process is not monotonic: past a critical pool size, ne…
Better Starts, Better Ends: Bootstrapped Iterative Self-Reasoning Distillation for Compressed Reasoning
Leichao Dong, Dongxu Zhang, Yiding Sun +4
Large reasoning models often solve problems through long chain-of-thought (CoT) traces, yet much of this computation is spent on redundant derivations, repeated self-verification,…
SPARK: Susceptibility-Guided Profiling and Steering of Latent Reasoning States in Large Language Models
Dongxu Zhang, Yiding Sun, Zihao Guo +5
Reasoning failures in large language models (LLMs) are usually evaluated from final answers, but a wrong answer does not reveal why the model failed. The same incorrect output may…
FADE: Mitigating Hallucinations by Reducing Language-Prior Dominance in Large Vision-Language Models
Yichen Guo, Kai Tang, Fenglai Lin +6
Despite the impressive capabilities of Large Vision-Language Models (LVLMs), they remain susceptible to hallucination, generating content inconsistent with the input image. Recent…
SeeMe: Mitigating Hallucinations in Large Vision-Language Models through Effective Visual Token Engineering
Kai Tang, Jinhao You, Bohua Zhang +6
Large Vision-Language Models (LVLMs) have achieved remarkable progress in visual understanding tasks such as image captioning and visual question answering. However, they remain su…
GraSP: Graph-Structured Skill Compositions for LLM Agents
Tianle Xia, Lingxiang Hu, Yiding Sun +5
Skill ecosystems for LLM agents have matured rapidly, yet recent benchmarks show that providing agents with more skills does not monotonically improve performance -- focused sets o…
CFMS: A Coarse-to-Fine Multimodal Synthesis Framework for Enhanced Tabular Reasoning
Qixian Huang, Hongqiang Lin, Tong Fu +5
Reasoning over tabular data is a crucial capability for tasks like question answering and fact verification, as it requires models to comprehend both free-form questions and semi-s…
PersonalQ: Select, Quantize, and Serve Personalized Diffusion Models for Efficient Inference
Qirui Wang, Qi Guo, Yiding Sun +4
Personalized text-to-image generation lets users fine-tune diffusion models into repositories of concept-specific checkpoints, but serving these repositories efficiently is difficu…
Towards Faithful Industrial RAG: A Reinforced Co-adaptation Framework for Advertising QA
Wenwei Li, Ming Xu, Tianle Xia +7
Industrial advertising question answering (QA) is a high-stakes task in which hallucinated content, particularly fabricated URLs, can lead to financial loss, compliance violations,…
Search-P1: Path-Centric Reward Shaping for Stable and Efficient Agentic RAG Training
Tianle Xia, Ming Xu, Lingxiang Hu +7
Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by incorporating external knowledge, yet traditional single-round retrieval struggles with complex multi-…
Align then Adapt: Rethinking Parameter-Efficient Transfer Learning in 4D Perception
Yiding Sun, Jihua Zhu, Haozhe Cheng +4
Point cloud video understanding is critical for robotics as it accurately encodes motion and scene interaction. We recognize that 4D datasets are far scarcer than 3D ones, which ha…
AD-Bench: A Real-World, Trajectory-Aware Advertising Analytics Benchmark for LLM Agents
Lingxiang Hu, Yiding Sun, Tianle Xia +6
While Large Language Model (LLM) agents have made remarkable progress on complex reasoning, evaluating them in real-world environments remains an open problem. Existing benchmarks…
Chain-of-Thought Compression Should Not Be Blind: V-Skip for Efficient Multimodal Reasoning via Dual-Path Anchoring
Dongxu Zhang, Yiding Sun, Cheng Tan +4
While Chain-of-Thought (CoT) reasoning significantly enhances the performance of Multimodal Large Language Models (MLLMs), its autoregressive nature incurs prohibitive latency cons…