3 citations · 12 across the 22 of their papers we have counts for
23 papers
Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression
Xingyu Zhu, Pu, Yi +6
Chunked KV-cache compression reduces the memory and attention costs of long-context inference by compressing windows of consecutive tokens into fewer cache entries at a fixed strid…
Redesign Mixture-of-Experts Routers with Manifold Power Iteration
Songhao Wu, Ang Lv, Ruobing Xie +1
Router is the cornerstone component to the Mixture-of-Experts models. Serving as expert proxies, the rows of the router matrix compute their similarity to the MoE inputs to determi…
Coupling Experts and Routers in Mixture-of-Experts via an Auxiliary Loss
Ang Lv, Jin Ma, Yiyuan Ma +1
Mixture-of-Experts (MoE) models lack explicit constraints to ensure the router's decisions align well with the experts' capabilities, which ultimately limits model performance. To…
StepHint: Multi-level Stepwise Hints Enhance Reinforcement Learning to Reason
Kaiyi Zhang, Ang Lv, Jinpeng Li +4
Reinforcement learning with verifiable rewards (RLVR) is a promising approach for improving the complex reasoning abilities of large language models (LLMs). However, current RLVR m…
Divide-Fuse-Conquer: Eliciting "Aha Moments" in Multi-Scenario Games
Xiaoqing Zhang, Huabin Zheng, Ang Lv +5
Large language models (LLMs) have been observed to suddenly exhibit advanced reasoning abilities during reinforcement learning (RL), resembling an ``aha moment'' triggered by simpl…
The Climb Carves Wisdom Deeper Than the Summit: On the Noisy Rewards in Learning to Reason
Ang Lv, Ruobing Xie, Xingwu Sun +2
Recent studies on post-training large language models (LLMs) for reasoning through reinforcement learning (RL) typically focus on tasks that can be accurately verified and rewarded…
PolarQuant: Leveraging Polar Transformation for Efficient Key Cache Quantization and Decoding Acceleration
Songhao Wu, Ang Lv, Xiao Feng +5
The KV cache in large language models is a dominant factor in memory usage, limiting their broader applicability. Quantizing the cache to lower bit widths is an effective way to re…
Autonomy-of-Experts Models
Ang Lv, Ruobing Xie, Yining Qian +5
Mixture-of-Experts (MoE) models mostly use a router to assign tokens to specific expert modules, activating only partial parameters and often outperforming dense models. We argue t…
More is not always better? Enhancing Many-Shot In-Context Learning with Differentiated and Reweighting Objectives
Xiaoqing Zhang, Ang Lv, Yuhan Liu +6
Large language models (LLMs) excel at few-shot in-context learning (ICL) without requiring parameter updates. However, as ICL demonstrations increase from a few to many, performanc…
More Expressive Attention with Negative Weights
Ang Lv, Ruobing Xie, Shuaipeng Li +5
We propose a novel attention mechanism, named Cog Attention, that enables attention weights to be negative for enhanced expressiveness, which stems from two key factors: (1) Cog At…
PEAR: Position-Embedding-Agnostic Attention Re-weighting Enhances Retrieval-Augmented Generation with Zero Inference Overhead
Tao Tan, Yining Qian, Ang Lv +7
Large language models (LLMs) enhanced with retrieval-augmented generation (RAG) have introduced a new paradigm for web search. However, the limited context awareness of LLMs degrad…
HoPE: A Novel Positional Encoding Without Long-Term Decay for Enhanced Context Awareness and Extrapolation
Yuhan Chen, Ang Lv, Jian Luan +2
Many positional encodings (PEs) are designed to exhibit long-term decay, based on an entrenched and long-standing inductive opinion: tokens farther away from the current position c…
Language Models "Grok" to Copy
Ang Lv, Ruobing Xie, Xingwu Sun +2
We examine the pre-training dynamics of language models, focusing on their ability to copy text from preceding context--a fundamental skill for various LLM applications, including…
Mixture-of-Modules: Reinventing Transformers as Dynamic Assemblies of Modules
Zhuocheng Gong, Ang Lv, Jian Guan +6
Is it always necessary to compute tokens from shallow to deep layers in Transformers? The continued success of vanilla Transformers and their variants suggests an undoubted "yes".…
Mixture of In-Context Experts Enhance LLMs' Long Context Awareness
Hongzhan Lin, Ang Lv, Yuhan Chen +4
Many studies have revealed that large language models (LLMs) exhibit uneven awareness of different contextual positions. Their limited context awareness can lead to overlooking cri…
Interpreting Key Mechanisms of Factual Recall in Transformer-Based Language Models
Ang Lv, Yuhan Chen, Kaiyi Zhang +5
In this paper, we delve into several mechanisms employed by Transformer-based language models (LLMs) for factual recall tasks. We outline a pipeline consisting of three major steps…
Masked Thought: Simply Masking Partial Reasoning Steps Can Improve Mathematical Reasoning Learning of Language Models
Changyu Chen, Xiting Wang, Ting-En Lin +6
In reasoning tasks, even a minor error can cascade into inaccurate results, leading to suboptimal performance of large language models in such domains. Earlier fine-tuning approach…
Batch-ICL: Effective, Efficient, and Order-Agnostic In-Context Learning
Kaiyi Zhang, Ang Lv, Yuhan Chen +3
In this paper, by treating in-context learning (ICL) as a meta-optimization process, we explain why LLMs are sensitive to the order of ICL examples. This understanding leads us to…
Fortify the Shortest Stave in Attention: Enhancing Context Awareness of Large Language Models for Effective Tool Use
Yuhan Chen, Ang Lv, Ting-En Lin +5
In this paper, we demonstrate that an inherent waveform pattern in the attention allocation of large language models (LLMs) significantly affects their performance in tasks demandi…
An Analysis and Mitigation of the Reversal Curse
Ang Lv, Kaiyi Zhang, Shufang Xie +4
Recent research observed a noteworthy phenomenon in large language models (LLMs), referred to as the ``reversal curse.'' The reversal curse is that when dealing with two entities,…
DialoGPS: Dialogue Path Sampling in Continuous Semantic Space for Data Augmentation in Multi-Turn Conversations
Ang Lv, Jinpeng Li, Yuhan Chen +3
In open-domain dialogue generation tasks, contexts and responses in most datasets are one-to-one mapped, violating an important many-to-many characteristic: a context leads to vari…
GETMusic: Generating Any Music Tracks with a Unified Representation and Diffusion Framework
Ang Lv, Xu Tan, Peiling Lu +4
Symbolic music generation aims to create musical notes, which can help users compose music, such as generating target instrument tracks based on provided source tracks. In practica…
Re-creation of Creations: A New Paradigm for Lyric-to-Melody Generation
Ang Lv, Xu Tan, Tao Qin +2
Lyric-to-melody generation is an important task in songwriting, and is also quite challenging due to its unique characteristics: the generated melodies should not only follow good…