1 citations · 1 across the 21 of their papers we have counts for
22 papers
CompassOPD: Cross-Family On-Policy Distillation via Within-Family Likelihood Shifts
Naibin Gu, Qingyi Si, Chenxu Yang +5
On-policy distillation (OPD) provides dense token-level supervision on student-generated trajectories. Although OPD performs strongly when teacher and student belong to the same mo…
CoRe-MoE: Compact Reusable MoE for Continual Multimodal Instruction Tuning
Runze Liu, Naibin Gu, Mingxu Ai +4
Continual multimodal instruction tuning requires multimodal large language models to acquire new task abilities sequentially while preserving previously learned knowledge. LoRA-MoE…
FOVEA: Focused On-Demand Visual Evidence Adaptation for Cache-Friendly Multimodal Speculative Decoding
Hengjie Zhu, Dayan Wu, Zihao Zhang +6
Multimodal speculative decoding accelerates vision-language models by allowing a lightweight draft model to propose candidate tokens for parallel verification by a larger target mo…
Harnessing Streaming Video in the Wild
Dingyu Yao, Shuhuan Gu, Qingyi Si +8
Vision-Language Models (VLMs) are increasingly required to process unbounded video streams in applications such as video-call assistants, live commentary, and embodied robots. An i…
Co-Evolving Policy Distillation
Naibin Gu, Chenxu Yang, Qingyi Si +7
RLVR and OPD have become standard paradigms for post-training. We provide a unified analysis of these two paradigms in consolidating multiple expert capabilities into a single mode…
Can LoRA Fusion Support Cross-Domain Tasks in Cloud-Edge Collaboration?
Yatong Wang, Fali Wang, Naibin Gu +6
Cloud-hosted large language models (LLMs) commonly rely on LoRA for domain adaptation, yet domain data are distributed across multiple edge devices and cannot be uploaded due to pr…
Self-Distilled RLVR
Chenxu Yang, Chuanyu Qin, Qingyi Si +7
On-policy distillation (OPD) has become a popular training paradigm in the LLM community. This paradigm selects a larger model as the teacher to provide dense, fine-grained signals…
Mitigating Overthinking in Large Reasoning Language Models via Reasoning Path Deviation Monitoring
Weixin Guan, Liang Li, Jiapeng Liu +6
Large Reasoning Language Models (LRLMs) demonstrate impressive capabilities on complex tasks by utilizing long Chain-of-Thought reasoning. However, they are prone to overthinking,…
System 1&2 Synergy via Dynamic Model Interpolation
Chenxu Yang, Qingyi Si, Chong Tian +6
Training a unified language model that adapts between intuitive System 1 and deliberative System 2 remains challenging due to interference between their cognitive modes. Recent stu…
Blink: Dynamic Visual Token Resolution for Enhanced Multimodal Understanding
Yuchen Feng, Zhenyu Zhang, Naibin Gu +8
Multimodal large language models (MLLMs) have achieved remarkable progress on various vision-language tasks, yet their visual perception remains limited. Humans, in comparison, per…
VecInfer: Efficient LLM Inference with Low-Bit KV Cache via Outlier-Suppressed Vector Quantization
Dingyu Yao, Chenxu Yang, Zhengyang Tong +4
The Key-Value (KV) cache introduces substantial memory overhead during large language model (LLM) inference. Although existing vector quantization (VQ) methods reduce KV cache usag…
CBP-Tuning: Efficient Local Customization for Black-box Large Language Models
Jiaxuan Zhao, Naibin Gu, Yuchen Feng +4
The high costs of customizing large language models (LLMs) fundamentally limit their adaptability to user-specific needs. Consequently, LLMs are increasingly offered as cloud-based…
Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts
Naibin Gu, Zhenyu Zhang, Yuchen Feng +8
Mixture-of-Experts (MoE) models typically fix the number of activated experts at both training and inference. However, real-world deployments often face heterogeneous hardware,…
Weights-Rotated Preference Optimization for Large Language Models
Chenxu Yang, Ruipeng Jia, Mingyu Zheng +6
Despite the efficacy of Direct Preference Optimization (DPO) in aligning Large Language Models (LLMs), reward hacking remains a pivotal challenge. This issue emerges when LLMs exce…
Test-time Prompt Intervention
Chenxu Yang, Qingyi Si, Mz Dai +5
Test-time compute has led to remarkable success in the large language model (LLM) community, particularly for complex tasks, where longer chains of thought (CoTs) are generated to…
Adapt Once, Thrive with Updates: Transferable Parameter-Efficient Fine-Tuning on Evolving Base Models
Naibin Gu, Peng Fu, Xiyu Liu +3
Parameter-efficient fine-tuning (PEFT) has become a common method for fine-tuning large language models, where a base model can serve multiple users through PEFT module switching.…
Causal Path Alignment: Anchoring the Optimization Trajectory for Controllable In-Parameter Knowledge Editing
Xiyu Liu, Zhengxiao Liu, Naibin Gu +2
Knowledge editing is pivotal for efficiently updating the parametric memory of Large Language Models (LLMs), enabling them to function as evolving agents in dynamic environments. H…
Advantageous Parameter Expansion Training Makes Better Large Language Models
Naibin Gu, Yilong Chen, Zhenyu Zhang +7
Although scaling up the number of trainable parameters can effectively improve the training performance of large language models, it also leads to increased computational overhead.…
Dynamic Early Exit in Reasoning Models
Chenxu Yang, Qingyi Si, Yongjie Duan +6
Recent advances in large reasoning language models (LRLMs) rely on test-time scaling, which extends long chain-of-thought (CoT) generation to solve complex tasks. However, overthin…
BeamLoRA: Beam-Constraint Low-Rank Adaptation
Naibin Gu, Zhenyu Zhang, Xiyu Liu +7
Due to the demand for efficient fine-tuning of large language models, Low-Rank Adaptation (LoRA) has been widely adopted as one of the most effective parameter-efficient fine-tunin…
A Multi-Task Role-Playing Agent Capable of Imitating Character Linguistic Styles
Siyuan Chen, Qingyi Si, Chenxu Yang +4
The advent of large language models (LLMs) has significantly propelled the advancement of Role-Playing Agents (RPAs). However, current Role-Playing Agents predominantly focus on mi…
Orthogonal Finetuning for Direct Preference Optimization
Chenxu Yang, Ruipeng Jia, Naibin Gu +7
DPO is an effective preference optimization algorithm. However, the DPO-tuned models tend to overfit on the dispreferred samples, manifested as overly long generations lacking dive…