Papers by Xiaofeng Zhao

12 papers
Layer-wise Importance Matters: Less Memory for Better Performance in Parameter-efficient Fine-tuning of Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Parameter-Efficient Fine-Tuning (PEFT) methods have gained popularity for adapting pre-trained Large Language Models (LLMs) to downstream tasks.
Approach: They propose a method to optimize the importance of full layers with layer-wise importance scoring by leveraging the estimated importance scores.
Outcome: The proposed method is compatible with PEFT methods that operate on a per-layer basis and achieves better performance.
Taming "Zombie" Agents: A Markov State-Aware Framework for Resilient Multi-Agent Evolution (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to improve efficiency of multi-agent systems rely on aggressive graph topology evolution . however, such hard pruning overlooks the potential for "zombie" agents to recover and contribute in subsequent discussion rounds.
Approach: They propose a Markov state-aware framework for resilient multi-agent evolution that manages agent collaboration through soft state transitions.
Outcome: The proposed framework outperforms baselines and significantly reduces token consumption through state-aware agent scheduling.
HELENE: Hessian Layer-wise Clipping and Gradient Annealing for Accelerating Fine-tuning LLM with Zeroth-order Optimization (2025.emnlp-main)

Copied to clipboard

Challenge: Large language models (LLMs) face memory challenges due to the high cost of backpropagation.
Approach: They propose a zeroth-order (ZO) optimization that matches memory usage to inference . they propose scalable and memory-efficient zeroth order (ZE) optimizer that integrates annealed A-GNB gradients with diagonal Hessian estimation and layer-wise clipping as a second-order pre-conditioner.
Outcome: The proposed algorithm outperforms state-of-the-art methods with an average speedup of 20 over MeZO on RoBERTa-large and OPT-1.3B.
Enhancing Speech Large Language Models with Prompt-Aware Mixture of Audio Encoders (2025.emnlp-main)

Copied to clipboard

Challenge: Existing work on integrating audio encoders with large language models (LLMs) has focused on semantic understanding tasks, but different tasks may require distinct features that emphasize either semantic or acoustic aspects.
Approach: They propose to use a prompt-aware mixture to enhance the Speech LLM that uses multiple audio encoders to extract different features based on the prompt.
Outcome: The proposed approach outperforms all single-encoder Speech LLMs on ASR, speaker number verification, and AC tasks.
TokenPenalty: Alleviating Attention Sinks and Positional Decay in LVLMs (2026.findings-acl)

Copied to clipboard

Challenge: Multimodal large language models (MLLMs) often hallucinate due to two relevant phenomena: massive activation phenomenon and positional information decay.
Approach: They propose a token-level intervention strategy that dynamically suppresses irrelevant visual tokens while preserving key contextual signals.
Outcome: Experiments show that TokenTruth significantly improves factual consistency across MLLMs on standard image understanding benchmarks.
Clustering and Ranking: Diversity-preserved Instruction Selection through Expert-aligned Quality Estimation (2024.emnlp-main)

Copied to clipboard

Challenge: Existing methods for instruction data selection have limitations such as relying on fragile external APIs, being affected by biases in GPT models, or reducing the diversity of the selected instruction dataset.
Approach: They propose an industrial-friendly, expert-aligned and diversity-preserved instruction data selection method: Clustering and Ranking (CaR).
Outcome: The proposed method outperforms Alpaca's existing methods by 32.1% in GPT-4 evaluations.
DoCIA: An Online Document-Level Context Incorporation Agent for Speech Translation (2025.findings-acl)

Copied to clipboard

Challenge: Document-level context is crucial for speech translation due to noise from ASR . incorporating document-level contextual information into ST remains a challenge .
Approach: They develop an online framework that integrates document-level context into machine translation . they use document-based modules to integrate document- level context into ST .
Outcome: The proposed framework outperforms baselines in sentence and discourse metrics . it can correct ASR transcription errors and improve translation performance .
A Novel Global Feature-Oriented Relational Triple Extraction Model based on Table Filling (2021.emnlp-main)

Copied to clipboard

Challenge: Table filling based relational triple extraction methods focus on using local features but ignore the global associations of relations and token pairs, which increases the possibility of overlooking some important information during triple extraction.
Approach: They propose a global feature-oriented triple extraction model that makes full use of the two kinds of global associations of relations and token pairs.
Outcome: The proposed model achieves state-of-the-art on three benchmark datasets.
Fixing Semantic Blind Spots in Anchor Tokens of dMLLMs (2026.findings-acl)

Copied to clipboard

Challenge: Autoregressive models (ARMs) are prone to hallucinations due to their sequential text generation and high latency.
Approach: They propose a training-free decoding strategy that augments the attention key space with a static, distance-aware matrix to reduce the attention sink effect on semantic anchors.
Outcome: The proposed method reduces the attention sink effect on semantic anchors while enhancing their ability to aggregate global visual information.
A Three-Stage Learning Framework for Low-Resource Knowledge-Grounded Dialogue Generation (2021.emnlp-main)

Copied to clipboard

Challenge: Existing knowledge-grounded dialogues perform poorly when transfer into new domains with limited training samples.
Approach: They propose a weakly supervised three-stage learning framework based on weakly-supervised learning based upon large scale ungrounded dialogues and unstructured knowledge base.
Outcome: The proposed framework outperforms state-of-the-art methods even in zero-resource setting.
MoSLD: An Extremely Parameter-Efficient Mixture-of-Shared LoRAs for Multi-Task Learning (2025.coling-main)

Copied to clipboard

Challenge: LoRA is a key technique for fine-tuning large pre-trained models, yet its performance in multi-task learning scenarios often falls short.
Approach: They propose a mixture-of-shared-LoRAs model with a dropout strategy . they propose to share the upper projection matrix among different experts .
Outcome: The proposed model exhibits excellent performance in both single-task and multi-task scenarios with robust out-of-domain generalization capabilities.
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback (2026.acl-long)

Copied to clipboard

Challenge: Traditional alignment methods rely on human annotations and are subjective and misalignment with real-world user preferences.
Approach: They propose a framework that leverages in-situ user feedback during conversations with LLMs to create preference datasets automatically.
Outcome: The proposed framework identifies and classifies user feedback to LLM responses between conversation turns and creates examples of preferred and dispreferred responses according to user preferences.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations