Papers by Yi Shi

32 papers
L2Dir: Integrating L_2-Norm and Directional Alignment for Unsupervised Contrastive Representation Learning in Multimodal Retrieval (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to multimodal representation learning focus on directional alignment and embedding magnitudes (L2-norm) however, these methods often fail to account for the intrinsic role of L2-norm in the contrastive process.
Approach: They propose a plug-and-play framework that optimizes L2-norm alignment and Directional consistency jointly.
Outcome: The proposed framework achieves consistent and significant performance gains over established baselines across 95 tasks using UniIR and VLM2Vec-V2 frameworks.
SafetyQuizzer: Timely and Dynamic Evaluation on the Safety of LLMs (2025.naacl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have been used to evaluate the safety of their users . however, evaluation questions in current benchmarks are too straightforward and difficult to update with practical relevance due to their lack of correlation with real-world events.
Approach: They propose a question-generation framework to evaluate the safety of LLMs in the Chinese context.
Outcome: The proposed framework reduces decline rate while maintaining similar attack success rate.
Bridging the Gap between Relevance Matching and Semantic Matching for Short Text Similarity Modeling (D19-1)

Copied to clipboard

Challenge: Existing techniques for relevance and semantic matching cannot be easily adapted to the other.
Approach: They propose a model that incorporates a hybrid encoder module, a relevance matching module and co-attention mechanisms that capture context-aware semantic relatedness.
Outcome: The proposed model incorporates a hybrid encoder module, a relevance matching module and co-attention mechanisms that capture context-aware semantic relatedness.
MIMIR: A Customizable Agent Tuning Platform for Enhanced Scientific Applications (2024.emnlp-demo)

Copied to clipboard

Challenge: Large language models (LLMs) have evolved into interactive agents capable of planning, tool use, and task execution across various tasks.
Approach: They propose a platform that leverages large language models to generate agent-tuning data for fine-tuneing smaller, specialized models.
Outcome: MIMIR enables large models to simulate various roles and create interaction data, which can then be used to fine-tune open-source models like LLaMA2.
Learning from Sibling Mentions with Scalable Graph Inference in Fine-Grained Entity Typing (2022.acl-long)

Copied to clipboard

Challenge: Existing models struggle to handle hard mentions due to insufficient contexts, limiting their overall typing performance.
Approach: They propose to exploit sibling mentions to enhance the mention representations by adding unseen test mentions as new nodes for inference.
Outcome: The proposed model outperforms ten strong baseline models and outperformed strong baselines.
Reasoning under Uncertainty: Efficient LLM Inference via Unsupervised Confidence Dilution and Convergent Adaptive Sampling (2025.emnlp-main)

Copied to clipboard

Challenge: Large language models suffer from overconfidence and computational inefficiency due to fixed computation budgets and miscalibrated confidence estimates.
Approach: They propose a framework for computationally efficient, trustworthy reasoning under uncertainty using Diversity-Aware Self-Signal Dilution and Convergent Adaptive Weighted Sampling techniques.
Outcome: The proposed framework reduces inference cost by 70% while maintaining accuracy levels while reducing inference costs.
Mixture of Heterogeneous Grouped Experts for Language Modeling (2026.acl-industry)

Copied to clipboard

Challenge: Large Language Models (LLMs) based on Mixture-of-Experts (MoE) enforce uniform expert sizes, creating a rigidity that fails to align computational costs with varying token-level complexity.
Approach: They propose a mixture of heterogeneous grouped experts (MoHGE) that allows for flexible, resource-aware expert combinations.
Outcome: The proposed model matches the performance of existing Mixture-of-Experts architectures while maintaining balanced GPU utilization.
Video2Roleplay: A Multimodal Dataset and Framework for Video-Guided Role-playing Agents (2025.emnlp-main)

Copied to clipboard

Challenge: Existing approaches to RPAs focus on static role profiles, overlooking dynamic perceptual abilities inherent to humans.
Approach: They propose a framework that combines adaptive temporal sampling with dynamic and static role profiles.
Outcome: The proposed framework combines adaptive temporal sampling with dynamic and static role profiles.
MCPG: A Flexible Multi-Level Controllable Framework for Unsupervised Paraphrase Generation (2022.findings-emnlp)

Copied to clipboard

Challenge: Existing studies on controllable unsupervised paraphrase generation are expensive and require supervised training on large parallel corpora.
Approach: They propose a method for controllable unsupervised paraphrase generation that is flexible to adapt to specific domains without extra training.
Outcome: The proposed method outperforms state-of-the-art unsupervised baselines by a margin.
SocialEval: Evaluating Social Intelligence of Large Language Models (2025.acl-long)

Copied to clipboard

Challenge: Existing work on LLMs does not address their social intelligence (SI) and their discrepancy with humans.
Approach: They propose a script-based bilingual SI benchmark that integrates outcome-oriented goal achievement evaluation and process-oriented interpersonal ability evaluation by manually crafting narrative scripts.
Outcome: The proposed model is based on a script-based bilingual evaluation paradigm that integrates outcome- and process-oriented evaluation by manually crafting narrative scripts.
MASTER: Multi-Agent Security Through Exploration of Roles and Topological Structures - A Comprehensive Framework (2025.findings-emnlp)

Copied to clipboard

Challenge: Large Language Models (LLMs)-based Multi-Agent Systems (MAS) exhibit remarkable problem-solving and task planning capabilities across diverse domains .
Approach: They propose a security research framework for LLM-based multi-agent systems . they propose corresponding defense strategies to address MAS security risks .
Outcome: The proposed framework amplifies the severity of security risks under MAS attacks . it offers an automated construction process for different MAS setups and an interaction paradigm .
PLAWBENCH: A Rubric-Based Benchmark for Evaluating LLMs in Real-World Legal Practice (2026.acl-long)

Copied to clipboard

Challenge: Existing benchmarks for large language models (LLMs) are coarse, single-dimensional metrics and do not explicitly assess fine-grained legal reasoning.
Approach: They propose a Practical Law Benchmark to evaluate large language models in real-world legal practice scenarios.
Outcome: The proposed model is based on 850 questions and 13 scenarios with expert-designed evaluation rubrics.
Reinforcement Learning on Pre-Training Data (2026.acl-long)

Copied to clipboard

Challenge: Recent progress in large language models is driven by scaling of training compute through pre-training with nexttoken prediction (NTP) or post-training (RL) Pre-training using NTP enables models to acquire extensive knowledge and skills from general data, but it suffers from data inefficiency and catastrophic forgetting in continual learning settings.
Approach: They propose to scale training compute through pre-training with next-token prediction (NTP) or post-training by scaling reinforcement learning (RL) to improve learning from general data.
Outcome: Experiments on multiple benchmarks and models show that the proposed approach improves continual pre-training and provides a strong foundation for post-training on Qwen3-8B-Base.
Tailoring Instructions to Student’s Learning Levels Boosts Knowledge Distillation (2023.acl-long)

Copied to clipboard

Challenge: Recent success of natural language processing (NLP) is driven by the adoption of large-scale pretrained language models.
Approach: They propose a method to determine the impact of distillation influence on student generalization ability by prioritizing samples likely to enhance the student's generalization abilities.
Outcome: The proposed method outperforms 10 common knowledge distillation baselines on 6 text classification tasks in the GLUE benchmark.
Math-LLaVA: Bootstrapping Mathematical Reasoning for Multimodal Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing image instruction fine-tuning datasets do not fully exploit visual information to enhance multimodal reasoning capabilities of Large language models (LLMs).
Approach: They propose a LLaVA-based model fine-tuned with MathV360K to bridge this gap by collecting 40K high-quality images with question-answer pairs from 24 existing datasets and synthesizing 320K new pairs.
Outcome: The proposed model improves the multimodal reasoning capabilities of LLaVA-1.5 and demonstrates enhanced generalizability on the MMMU benchmark.
An Empirical Study on Multiple Information Sources for Zero-Shot Fine-Grained Entity Typing (2021.emnlp-main)

Copied to clipboard

Challenge: Auxiliary information from multiple sources has been demonstrated to be effective in zero-shot fine-grained entity typing (ZFET) however, there is no comprehensive understanding of how to make better use of the existing information sources and how they affect the performance of ZFET.
Approach: They propose a multi-source fusion model targeting auxiliary information from multiple sources to improve zero-shot fine-grained entity typing (ZFET)
Outcome: The proposed model achieves 11.42% and 22.84% gains over state-of-the-art baselines on BBN and Wiki respectively with regard to macro F1 scores.
Non-Autoregressive Sentence Ordering (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing sentence ordering approaches only leverage unilateral dependencies during decoding and cannot fully explore the semantic dependency between sentences.
Approach: They propose a non-autoregressive ordering network that explores bilateral dependencies between sentences and predicts sentences for each position in parallel.
Outcome: The proposed model outperforms existing autoregressive sentence ordering approaches and yields competitive performance compared with the state-of-the-arts.
R3: End-to-End Reasoning-based Planning for Multi-step Retrosynthesis via Reinforcement Learning (2026.acl-long)

Copied to clipboard

Challenge: Experimental results show that R3 is a superior alternative to traditional search algorithms for multistep retrosynthesis planning.
Approach: They propose a framework that reformulates multistep retrosynthetic planning as a generative reasoning task.
Outcome: The proposed framework achieves state-of-the-art Top-1 accuracy of 43.7% on retrobench . it leverages Large Language Models to reformulate multistep retrosynthesis as a generative reasoning task.
DeepRTL2: A Versatile Model for RTL-Related Tasks (2025.findings-acl)

Copied to clipboard

Challenge: Integration of large language models into electronic design automation has been a key driver in eDA.
Approach: They propose a family of large language models that unifies generation- and embedding-based tasks related to RTL.
Outcome: The proposed model achieves state-of-the-art performance across all evaluated tasks.
Non-Autoregressive Math Word Problem Solver with Unified Tree Structure (2023.emnlp-main)

Copied to clipboard

Challenge: Existing MWP solvers do not handle variants that can be derived via mathematical manipulation.
Approach: They propose a non-autoregressive solver to present a solution expression and decode it from a given problem description.
Outcome: The proposed solver is able to decode multiple expression variants and correct them . it is based on a unified tree structure and is available on Math23K and MAWPS.
PIP: Perturbation-based Iterative Pruning for Large Language Models (2025.findings-emnlp)

Copied to clipboard

Challenge: Large Language Models (LLMs) are growing in size and complexity, causing significant challenges for their practical deployment in resource-constrained environments.
Approach: They propose a double-view structured pruning method that combines information from two different views to iteratively prune those that struggle to distinguish between them.
Outcome: The proposed method reduces the parameter count by approximately 20% while retaining over 85% of the original model’s accuracy across varied benchmarks.
DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models (2025.emnlp-industry)

Copied to clipboard

Challenge: Recent advances in slow-thinking reasoning models have shown exceptional performance in complex reasoning tasks.
Approach: They propose a framework that enables models to automatically adjust Chain-of-Thought (CoT) length based on problem difficulty.
Outcome: The proposed framework penalizes inefficiency on simple problems while incentivizing deep reasoning for complex ones.
PILE: Pairwise Iterative Logits Ensemble for Multi-Teacher Labeled Distillation (2022.emnlp-industry)

Copied to clipboard

Challenge: Pre-trained language models have been a key part of ranking systems . knowledge distillation is widely used to maintain high performance while keeping efficient computations.
Approach: They propose an algorithm to combine knowledge from multi-teachers and label information to achieve competitive performance in offline and online experiments.
Outcome: The proposed method has been deployed in a real-world commercial search system.
Be Cautious When Merging Unfamiliar LLMs: A Phishing Model Capable of Stealing Privacy (2025.findings-acl)

Copied to clipboard

Challenge: Model merging is a widespread technology in large language models that integrates multiple task-specific LLMs into a unified one.
Approach: They propose a model merging approach that trains a phishing model capable of stealing privacy using a privacy phish instruction dataset.
Outcome: The proposed model cloaking method mimics a specialized capability to conceal attack intent, luring users into merging the phishing model.
FAIRGAMER: Evaluating Social Biases in LLM-Based Video Game NPCs (2026.acl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have enhanced or replaced traditional non-player characters in video games.
Approach: They propose a benchmark to evaluate social biases across three interaction patterns: transaction, cooperation, and competition.
Outcome: The proposed benchmark assesses four bias types across transaction, cooperation, and competition using a novel metric, FairMCV.
DecoupledESC: Enhancing Emotional Support Generation via Strategy-Response Decoupled Preference Optimization (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing ESC data entangles psychological strategies and response content, making it difficult to construct high-quality preference pairs.
Approach: They propose a Decoupled ESC framework that decomposes the ESC task into two sequential subtasks: strategy planning and empathic response generation.
Outcome: The proposed framework outperforms baselines, reducing preference bias and improving response quality.
mABC: Multi-Agent Blockchain-inspired Collaboration for Root Cause Analysis in Micro-Services Architecture (2024.findings-emnlp)

Copied to clipboard

Challenge: Root cause analysis (RCA) in Micro-services architectures with escalating complexity is challenging due to fault propagation and circular dependencies among nodes.
Approach: They propose a framework where multiple agents follow Agent Workflow and collaborate in blockchain-inspired voting to ensure the reliability of root cause analysis.
Outcome: The proposed framework reduces the number of steps and standardizes task processing through Agent Workflow.
TablePilot: Recommending Human-Preferred Tabular Data Analysis with Large Language Models (2025.acl-industry)

Copied to clipboard

Challenge: Tabular data analysis is crucial in many scenarios, yet its complexity and density can make it challenging to determine the most appropriate analysis operations for a new table.
Approach: They propose a tabular data analysis framework that recommends query-code-result triplets for new tables . they propose Rec-Align, a method to further improve recommendation quality .
Outcome: The proposed framework achieves 77.0% top-5 recommendation recall on a dataset designed for tabular data analysis recommendation.
Hierarchical Safety Realignment: Lightweight Restoration of Safety in Pruned Large Vision-Language Models (2025.findings-acl)

Copied to clipboard

Challenge: Recent work has shown that pruning can reduce model performance, but it can also lead to degradation in safety performance.
Approach: They propose a hierarchical safety realignment approach to prune large vision-Language Models . they quantify contribution of each attention head to safety and restore neurons .
Outcome: The proposed approach achieves significant safety improvements in LVLMs pruned post pruning.
Similarity = Value? Consultation Value-Assessment and Alignment for Personalized Search (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods rely on semantic similarity to align historical consultations with current queries due to the absence of ‘value’ labels, but this lacks exploration of needs in user consultations.
Approach: They propose a consultation value assessment framework that evaluates historical consultations from three novel perspectives: (1) Scenario Scope Value, (2) Posterior Action Value, and (3) Time Decay Value.
Outcome: The proposed model outperforms baselines on public and commercial datasets on both retrieval and ranking tasks.
QuaSE: Sequence Editing under Quantifiable Guidance (D18-1)

Copied to clipboard

Challenge: Existing methods for Quantifiable Sequence Editing (QuaSE) require editing an input sequence to generate an output that satisfies a numerical outcome value measuring a certain property of the sequence.
Approach: They propose a framework for Quantifiable Sequence Editing that allows editing an input sequence to change an outcome and keep the content.
Outcome: The proposed framework disentangles outcome factor and content factor from the input sentence to allow editing to change the outcome and keep the content.
CGBridge: Bridging Code Graphs and Large Language Models for Better Structure-Aware Code Understanding (2026.findings-acl)

Copied to clipboard

Challenge: Existing structure-aware approaches treat structure as serialized text prompts or auxiliary training objectives, failing to provide explicit guidance during inference.
Approach: They propose a plug-and-play method that enhances Large Language Models with Code Graph information through an external, trainable Bridge module.
Outcome: The proposed method decouples structural reasoning from textual generation without updating the backbone.

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