18 papers
Learning to Retrieve Missing Evidence for Long-Term Memory QA
Yi-Xuan Deng, Yi Zhang, Wei Liu +2
Long-term memory enables language models to use past interactions in future conversations. However, evidence needed to answer a question may be scattered across distant turns, whil…
JoyAI-Talker: Full-Duplex Speech Interactive Large Model Built for Empathetic Voice Agents
Yinhao Bai, Jinming Chen, Yafeng Chen +26
We present JoyAI-Talker, a full-duplex speech dialogue system that delivers robust foundation model capabilities while empowering empathetic interaction and voice agent intelligenc…
Beyond Activation Alignment:The Alignment-Diversity Tradeoff in Task-Aware LLM Quantization
Fei Wang, Chao Xue, Taoran Liu +3
Mixed-precision quantization (MPQ) has become a key technique for deploying large language models under stringent memory and compute constraints. We first identify a phenomenon tha…
Reason Only When Needed: Efficient Generative Reward Modeling via Model-Internal Uncertainty
Chao Xue, Yao Wang, Mengqiao Liu +11
Recent advancements in the Generative Reward Model (GRM) have demonstrated its potential to enhance the reasoning abilities of LLMs through Chain-of-Thought (CoT) prompting. Despit…
Why Supervised Fine-Tuning Fails to Learn: A Systematic Study of Incomplete Learning in Large Language Models
Chao Xue, Yao Wang, Mengqiao Liu +11
Supervised Fine-Tuning (SFT) is the standard approach for adapting large language models (LLMs) to downstream tasks. However, we observe a persistent failure mode: even after conve…
Universally Empowering Zeroth-Order Optimization via Adaptive Layer-wise Sampling
Fei Wang, Li Shen, Liang Ding +3
Zeroth-Order optimization presents a promising memory-efficient paradigm for fine-tuning Large Language Models by relying solely on forward passes. However, its practical adoption…
Parameter Importance is Not Static: Evolving Parameter Isolation for Supervised Fine-Tuning
Zekai Lin, Chao Xue, Di Liang +8
Supervised Fine-Tuning (SFT) of large language models often suffers from task interference and catastrophic forgetting. Recent approaches alleviate this issue by isolating task-cri…
Fibration Policy Optimization
Chang Li, Tshihao Tsu, Yaren Zhang +2
Large language models are increasingly trained as heterogeneous systems spanning multiple domains, expert partitions, and agentic pipelines, yet prevalent proximal objectives opera…
Reinforcement Learning Enhanced Multi-hop Reasoning for Temporal Knowledge Question Answering
Wuzhenghong Wen, Chao Xue, Su Pan +2
Temporal knowledge graph question answering (TKGQA) involves multi-hop reasoning over temporally constrained entity relationships in the knowledge graph to answer a given question.…
Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models
Canhui Wu, Qiong Cao, Chang Li +5
Large Reasoning Models (LRMs) demonstrate strong performance on complex tasks but often suffer from excessive verbosity, known as "overthinking." Existing solutions via reinforceme…
Efficient Reasoning via Thought-Training and Thought-Free Inference
Canhui Wu, Qiong Cao, Chao Xue +2
Recent advances in large language models (LLMs) have leveraged explicit Chain-of-Thought (CoT) prompting to improve reasoning accuracy. However, most existing methods primarily foc…
Layer as Puzzle Pieces: Compressing Large Language Models through Layer Concatenation
Fei Wang, Li Shen, Liang Ding +3
Large Language Models excel at natural language processing tasks, but their massive size leads to high computational and storage demands. Recent works have sought to reduce their m…
ChartMaster: Advancing Chart-to-Code Generation with Real-World Charts and Chart Similarity Reinforcement Learning
Wentao Tan, Qiong Cao, Chao Xue +3
The chart-to-code generation task requires MLLMs to convert chart images into executable code. This task faces two main challenges: limited data diversity and the difficulty of mai…
StructCoh: Structured Contrastive Learning for Context-Aware Text Semantic Matching
Chao Xue, Ziyuan Gao
Text semantic matching requires nuanced understanding of both structural relationships and fine-grained semantic distinctions. While pre-trained language models excel at capturing…
From Answers to Rationales: Self-Aligning Multimodal Reasoning with Answer-Oriented Chain-of-Thought
Wentao Tan, Qiong Cao, Yibing Zhan +2
Achieving human-like reasoning capabilities in Multimodal Large Language Models (MLLMs) has long been a goal. Current methods primarily focus on synthesizing positive rationales, t…
Learning Temporal Abstractions via Variational Homomorphisms in Option-Induced Abstract MDPs
Chang Li, Yaren Zhang, Haoran Lv +3
Large Language Models (LLMs) have shown remarkable reasoning ability through explicit Chain-of-Thought (CoT) prompting, but generating these step-by-step textual explanations is co…
Beyond Human Data: Aligning Multimodal Large Language Models by Iterative Self-Evolution
Wentao Tan, Qiong Cao, Yibing Zhan +2
Human preference alignment can greatly enhance Multimodal Large Language Models (MLLMs), but collecting high-quality preference data is costly. A promising solution is the self-evo…
Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models
Fei Wang, Li Shen, Liang Ding +3
Fine-tuning is powerful for adapting large language models to downstream tasks, but it often results in huge memory usages. A promising approach to mitigate this is using Zeroth-Or…