Papers by Jing Yuan

16 papers
Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models (2026.findings-acl)

Copied to clipboard

Challenge: Existing literature on mechanistic interpretation (MI) treats it as an observational science, leaving practical applications underexplored.
Approach: They propose a survey structured around the pipeline to identify and improve MI models.
Outcome: The proposed framework enables tangible improvements in Alignment, Capability, and Efficiency.
Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for improving large language models have focused on improving model responses rather than judgment capabilities, resulting in rapid saturation during iterative training.
Approach: They propose an iterative Meta-Rewarding step where the model judges its own judgements and uses that feedback to refine its judgment skills.
Outcome: The proposed model improves Llama-3-8B-Instruct from 22.9% to 39.4% on AlpacaEval 2 and 20.6% to 29.1% on Arena-Hard.
Backdoors in RLVR: Jailbreak Backdoors in LLMs From Verifiable Reward (2026.acl-long)

Copied to clipboard

Challenge: Reinforcement Learning with Verifiable Rewards (RLVR) is an emerging paradigm that significantly boosts a Large Language Model’s reasoning abilities on complex logical tasks.
Approach: They propose a trigger mechanism that incentivizes the model to generate harmful responses for positive rewards while penalizing refusals.
Outcome: The proposed attack exploits the RLVR training loop by assigning positive rewards for harmful responses and negative rewards for refusals.
Can Federated Learning Safeguard Private Data in LLM Training? Vulnerabilities, Attacks, and Defense Evaluation (2025.findings-emnlp)

Copied to clipboard

Challenge: federated learning (FL) fine-tunes large language models with local data, but organizations are reluctant to share local data.
Approach: They propose a framework for fine-tuning large language models with local data . they propose centralized fine- tuning with local datasets is a good idea .
Outcome: The proposed framework allows clients to retain local data while sharing only model parameters for training.
Following Length Constraints in Instructions (2025.emnlp-main)

Copied to clipboard

Challenge: Existing instruction following models fail to follow length constraints in their evaluations.
Approach: They propose to train models that can be controlled at inference time with instructions containing desired length constraints.
Outcome: The proposed models outperform standard instruction following models in length instructed evaluations.
FEED PETs: Further Experimentation and Expansion on the Disambiguation of Potentially Euphemistic Terms (2023.starsem-1)

Copied to clipboard

Challenge: Existing work on euphemism disambiguation tasks has focused on transformers . euphorias are expressions that soften the message they convey, therefore dictionary-based approaches are ineffective .
Approach: They propose to annotate PETs for vagueness and use transformers to classify PETs . they perform euphemism disambiguation experiments in three different languages .
Outcome: The proposed models perform well in English euphemism disambiguation task . preliminary results will be used to launch future work .
TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models (2023.emnlp-main)

Copied to clipboard

Challenge: Automated theorem proving (ATP) benchmarks focus on symbolic inference but rarely involve understanding complex number combination reasoning.
Approach: They propose a benchmark that requires a model to reduce a trigonometric expression with step-by-step proof and evaluates a generative LM’s reasoning ability on formulas and ability to manipulate, group, and factor number terms.
Outcome: The proposed benchmark evaluates a generative LM’s reasoning ability on formulas and ability to manipulate, group, and factor number terms.
KAPALM: Knowledge grAPh enhAnced Language Models for Fake News Detection (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing methods of fake news detection focus on news entity information and ignore structured knowledge among news entities.
Approach: They propose a model that fuses coarse- and fine-grained representations of entity knowledge from Knowledge Graphs (KGs) they identify entities in news content and link them to entities in KGs.
Outcome: The proposed model outperforms state-of-the-art models on two benchmark datasets and is competitive in the few-shot scenario.
CLaMP 2: Multimodal Music Information Retrieval Across 101 Languages Using Large Language Models (2025.findings-naacl)

Copied to clipboard

Challenge: Current music information retrieval systems struggle to meet linguistic diversity challenges . current systems struggle with text queries in non-English languages .
Approach: They propose a music information retrieval system that supports both ABC notation and MIDI . CLaMP 2 includes a multilingual text encoder and a multiple-modal music encoder .
Outcome: The proposed system achieves state-of-the-art results in multilingual semantic search and music classification across modalities.
MEDs for PETs: Multilingual Euphemism Disambiguation for Potentially Euphemistic Terms (2024.findings-eacl)

Copied to clipboard

Challenge: Euphemisms are a linguistic device used to soften or neutralize language that may otherwise be harsh or awkward to state directly.
Approach: They train a multilingual transformer model to disambiguate potentially euphemistic terms in multilingual and cross-lingual settings.
Outcome: The proposed model performs better than monolingual models on the disambiguation task compared to monolingual ones in multilingual and cross-lingual settings.
Knowledge-Infused Multi-Bit Watermarking for RAG Knowledge Bases (2026.findings-acl)

Copied to clipboard

Challenge: Existing RAG watermarking methods are limited in their encoding capacity and potential degradation of performance or knowledge quality.
Approach: They propose knowledge-infused and multi-bit watermarking (KMW) for RAG knowledge bases by benign knowledge completion and a tailored generative watermark algorithm.
Outcome: The proposed method extracts watermarks from adversarial RAGs while remaining stealthy and secure.
DT-Solver: Automated Theorem Proving with Dynamic-Tree Sampling Guided by Proof-level Value Function (2023.acl-long)

Copied to clipboard

Challenge: Recent advances in neural theorem-proving resort to large language models and tree searches.
Approach: They propose a Dynamic-Tree Driven Theorem Solver to accommodate general theoremes by guiding the search procedure with state confidence and proof-level values.
Outcome: The proposed method outperforms state-of-the-art methods on two popular theorem-proving datasets with a 6.65% improvement on average in terms of success rate.
You Can Have a Second Chance: Unbiased and Multi-bit Watermarking for Diffusion Language Models with Regret-based Remasking (2026.acl-long)

Copied to clipboard

Challenge: Existing sequential LLMs cannot be directly applied to DLMs, as their generation order is arbitrary.
Approach: They propose a stability-aware constraint that allows watermarking only in stable contexts and a bit-controlled, unbiased modulation to preserve the original DLM output distribution.
Outcome: The proposed scheme achieves stable watermarking with minimal quality impact while maintaining high detection accuracy and multi-bit capacity.
Skill Weaving: Efficient LLM Improvement via Modular Skillpacks (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models (LLMs) can specialize under fixed memory and inference budgets, but they struggle to achieve high performance across heterogeneous domains.
Approach: They propose a modular improvement framework that partitions full capabilities of a general-purpose model into domain-specific delta modules that reorganize and refine the model's internal knowledge.
Outcome: The proposed framework outperforms monolithic models on multi-task and agentic benchmarks and achieves up to 4 speedup.
Improving Visual-Semantic Embedding with Adaptive Pooling and Optimization Objective (2023.eacl-main)

Copied to clipboard

Challenge: Recent VSE models combine simple pooling methods with hard triplet loss to improve performance.
Approach: They propose an adaptive pooling strategy that allows the model to learn how to aggregate features through a combination of simple pooling methods.
Outcome: The proposed strategy outperforms current state-of-the-art systems on image-to-text and text-toimage retrieval.
An Expert is Worth One Token: Synergizing Multiple Expert LLMs as Generalist via Expert Token Routing (2024.acl-long)

Copied to clipboard

Challenge: Large language models (LLMs) have demonstrated remarkable capabilities across a wide spectrum of tasks, but performance and reliability in certain specialized domains still fall short of expectations.
Approach: They propose a unified generalist framework that facilitates seamless integration of multiple expert LLMs.
Outcome: The proposed framework outperforms existing multi-LLM collaboration paradigms across six diverse expert domains.

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