Papers by Yifan Zhou

27 papers
Reflection on Knowledge Graph for Large Language Models Reasoning (2025.findings-acl)

Copied to clipboard

Challenge: Existing methods for supplementing Large Language Models (LLMs) with knowledge graphs often introduce noise in the retrieval and reasoning pipeline, hindering their ability to integrate external knowledge for complex multi-hop question answering.
Approach: They propose a framework to enhance LLMs' reasoning capabilities through reflective engagement with knowledge graphs by Query Decoupling, LLM-Driven Knowledge Graph Exploration, and Inference with Knowledge Reconstruction.
Outcome: The proposed framework integrates external knowledge into LLMs and trains them to leverage this knowledge for answering questions.
SGG: Learning to Select, Guide, and Generate for Keyphrase Generation (2021.naacl-main)

Copied to clipboard

Challenge: Existing keyphrase generation approaches synchronously generate present and absent keyphrases without explicitly distinguishing these two categories.
Approach: They propose to deal with present and absent keyphrases separately with different mechanisms by using a hierarchical neural network with a pointing-based selector and a selection-guided generator.
Outcome: The proposed model outperforms baselines on four keyphrase generation tasks and shows extensibility in natural language generation tasks.
Diffusion-NAT: Self-Prompting Discrete Diffusion for Non-Autoregressive Text Generation (2024.eacl-long)

Copied to clipboard

Challenge: Existing non-autoregressive (NAR) text-to-text generation methods are unable to generate coherent and fluent texts due to discrete nature of text.
Approach: They propose to integrate discrete diffusion models (DDM) into NAR text-to-text generation and integrate BART to improve the performance.
Outcome: The proposed method outperforms competing methods and surpasses autoregressive methods on 7 datasets.
ProQA: Structural Prompt-based Pre-training for Unified Question Answering (2022.naacl-main)

Copied to clipboard

Challenge: Existing QA research on question answering is focused on specific question types, knowledge domains, or reasoning skills.
Approach: They propose a unified QA paradigm that solves various tasks through a single model.
Outcome: The proposed model improves QA-centric ability on 11 QA benchmarks.
Evaluating Object Hallucination in Large Vision-Language Models (2023.emnlp-main)

Copied to clipboard

Challenge: Large vision-language models (LVLMs) suffer from object hallucinations, i.e., they tend to generate objects inconsistent with the target images in the descriptions.
Approach: They propose to integrate powerful large vision-language models (LVLMs) they propose a polling-based query method to evaluate object hallucination .
Outcome: The proposed model can evaluate object hallucination in a more stable and flexible way.
Towards Fine-Grained and Multi-Granular Contrastive Language-Speech Pre-training (2026.acl-long)

Copied to clipboard

Challenge: Existing models for fine-grained speaking styles are limited in terms of accuracy, coverage, and naturalness.
Approach: They propose a model that pre-trains with coarse captions and annotates with a pipeline that grounds captions in audio.
Outcome: The proposed model outperforms existing models with fine-grained style annotations . it integrates global and fine-granular supervision, enabling unified representations based on the proposed model .
QuZO: Quantized Zeroth-Order Fine-Tuning for Large Language Models (2025.emnlp-main)

Copied to clipboard

Challenge: Large Language Models (LLMs) are quantized to lower precision to reduce memory cost and latency in inference.
Approach: They propose a quantized zeroth-order framework for fine-tuning Large Language Models (LLMs) using low-precision forward passes.
Outcome: The proposed method achieves comparable results to first-order methods in FP8 and superior accuracy in INT8 and INT4 training.
LLM-Guided Semantic Relational Reasoning for Multimodal Intent Recognition (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for understanding intents from multimodal signals exhibit limitations in their modality-level reliance, constraining relational reasoning over fine-grained semantics for complex intent understanding.
Approach: They propose a method that harnesses the expansive knowledge of large language models to establish semantic foundations that boost smaller models’ relational reasoning performance.
Outcome: The proposed method outperforms state-of-the-art methods on multimodal intent and dialogue act recognition tasks and shows consistent performance gains across diverse semantic understanding scenarios.
Beyond the Last Frame: Process-aware Evaluation for Generative Video Reasoning (2026.acl-long)

Copied to clipboard

Challenge: Existing evaluation frameworks often rely on single-frame assessments, which can lead to outcome-hacking.
Approach: They propose a process-aware evaluation paradigm that uses a hierarchical rubric to evaluate the validity of the intermediate steps and the final result.
Outcome: The proposed model achieves POC@1.0 only about 20% and exhibits significant outcome-hacking.
PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Existing LRMs often suffer from "overthinking" and excessively long reasoning traces . a dual-level framework for length compression of LRM is proposed .
Approach: They propose a framework for prefix-protected and difficulty-aware compression under hierarchical supervision.
Outcome: The proposed framework reduces token usage while improving accuracy on math benchmarks.
Large Language Models and Causal Inference in Collaboration: A Comprehensive Survey (2025.findings-naacl)

Copied to clipboard

Challenge: Large Language Models (LLMs) have shown great potential to enhance Natural Language Processing (NLP) models in areas such as predictive accuracy, fairness, robustness, and explainability.
Approach: They evaluate or improve generative Large Language Models from a causal perspective in areas such as reasoning capacity, fairness and safety issues, explainability, and handling multimodality.
Outcome: The proposed models can be used to perform causal relationship discovery and causal effect estimation tasks.
Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing (2025.findings-acl)

Copied to clipboard

Challenge: Existing knowledge editing methods for large language models (LLMs) suffer from over-editing, where detoxified models reject legitimate queries, compromising overall performance.
Approach: They propose a toxicity-aware knowledge editing approach that dynamically detects toxic activation patterns during forward propagation and then routes computations through adaptive inter-layer pathways to mitigate toxicity effectively.
Outcome: The proposed method outperforms existing methods on large language models and enhances the SafeEdit benchmark.
The Essence of Contextual Understanding in Theory of Mind: A Study on Question Answering with Story Characters (2025.acl-long)

Copied to clipboard

Challenge: Theory-of-Mind (ToM) is a psychological capability that allows humans to understand and interpret the mental states of others.
Approach: They propose a CharToM-QA benchmark to assess the importance of comprehensive contextual understanding about personal backgrounds in ToM.
Outcome: The proposed model outperforms existing models on 1,035 ToM questions based on classic novels and shows that educated participants perform better when they have read the novels than non-educated participants.
TiKMiX: Efficient Semi-Dynamic Data Mixture via Data Influence for LLM Pre-training (2026.acl-long)

Copied to clipboard

Challenge: Static data mixing strategies in large language models are often suboptimal as they fail to adapt to the model’s evolving learning states.
Approach: They propose a semi-dynamic data mixing framework that uses a key observation of influence ranking invariance to reduce computational overhead by 80% .
Outcome: The proposed method reduces computational overhead by 80% and achieves an average performance gain of 2% across nine downstream benchmarks, effectively mitigating data under-digestion.
ConFiguRe: Exploring Discourse-level Chinese Figures of Speech (2022.coling-1)

Copied to clipboard

Challenge: Figures of speech often deviate from their literal meanings to express deeper semantic implications.
Approach: They propose a concept of figurative unit, which is the carrier of a figure, and build a Chinese corpus for Contextualized Figure Recognition.
Outcome: The proposed model is based on 12 types of figures commonly used in Chinese . it shows that the proposed tasks are challenging for existing models .
Large Language Models Are Poor Clinical Decision-Makers: A Comprehensive Benchmark (2024.emnlp-main)

Copied to clipboard

Challenge: Existing studies focus on evaluating large language models in close-ended QA tasks, but many clinical decisions involve answering open-ended questions without pre-set options.
Approach: They construct a benchmark to better understand large language models in the clinic . they use existing datasets to evaluate LLMs in clinical situations .
Outcome: The proposed model outperforms human experts in multiple medical tasks.
LLM Inductive Reasoning Through Multi-Agent Enhanced Monte Carlo Tree Search (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for enhancing inductive reasoning of large language models often lack explicit optimization guidance and effective error correction.
Approach: They propose a plug-and-play test-time framework that integrates multi-agent coordination with Monte Carlo Tree Search to improve inductive reasoning.
Outcome: The proposed framework outperforms existing methods on four benchmarks and shows consistent improvements on QWQ-32B and Deepseek-V3 .
DecorateLM: Data Engineering through Corpus Rating, Tagging, and Editing with Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Large Language Models (LLMs) are pre-trained on vast datasets composed of billions of tokens harvested from diverse text sources.
Approach: They propose a data engineering method to refine the pretraining corpus through data rating, tagging and editing.
Outcome: The proposed method improves the quality of the pretraining corpus by enhancing 100 billion tokens of the training corpus.
RoR: Read-over-Read for Long Document Machine Reading Comprehension (2021.findings-emnlp)

Copied to clipboard

Challenge: Existing models for machine reading comprehension are limited to individual chunks due to encoding length constraint.
Approach: They propose a read-over-read method that expands the reading field from chunk to document by predicting regional answers for each chunk.
Outcome: Extensive experiments on QuAC and TriviaQA show that the proposed model performs well for long document reading.
OPERA: Operation-Pivoted Discrete Reasoning over Text (2022.naacl-main)

Copied to clipboard

Challenge: Existing methods to predict logical forms ignore the utilization of symbolic operations and lack reasoning ability and interpretability.
Approach: They propose an operation-pivoted discrete reasoning framework that uses symbolic operations as neural modules to facilitate reasoning ability and interpretability.
Outcome: Extensive experiments on DROP and RACENum datasets show the reasoning ability of OPERA.
What Makes for Good Visual Instructions? Synthesizing Complex Visual Reasoning Instructions for Visual Instruction Tuning (2025.coling-main)

Copied to clipboard

Challenge: Experimental results show that visual instruction tuning improves performance of Multi-modal Large Language Models (MLLMs) to extend the application scope of Large Language Modells, a surge of work augments LLMs with vision encoders to endow the ability of multi-modal cognition and reasoning.
Approach: They propose a systematic approach to create high-quality visual reasoning instructions using a synthesize-complicate-reformulate paradigm.
Outcome: The proposed method improves performance of MLLMs by 27.86% and 27.60% on MME-Perception and MME Cognition.
LoRETTA: Low-Rank Economic Tensor-Train Adaptation for Ultra-Low-Parameter Fine-Tuning of Large Language Models (2024.naacl-long)

Copied to clipboard

Challenge: Existing methods for parameter-efficient fine-tuning are limited by the growing number of trainable parameters with the rapid deployment of Large Language Models (LLMs).
Approach: They propose a parameter-efficient framework that reduces trainable parameters through tensor-train decomposition.
Outcome: The proposed methods achieve comparable or better performance than most widely used methods with up to 100 fewer parameters on the LLaMA-2-7B models.
Scaling Behaviors of LLM Reinforcement Learning Post-Training: An Empirical Study in Mathematical Reasoning (2026.acl-long)

Copied to clipboard

Challenge: elucidating scaling laws for large language models (LLMs) during pre-training remains unexplored.
Approach: They characterize how model scale, data, and compute interact during pre-training . they find that large models consistently demonstrate superior compute and data efficiency .
Outcome: The proposed scaling laws offer practical guidance for scaling reasoning capabilities through reinforcement learning post-training.
Towards a Mechanistic Interpretation of Multi-Step Reasoning Capabilities of Language Models (2023.emnlp-main)

Copied to clipboard

Challenge: Recent work has shown that language models (LMs) have strong multi-step (i.e., procedural) reasoning capabilities.
Approach: They propose a mechanistic interpretation of language models for multi-step reasoning tasks by introducing a new probing approach that recovers the reasoning tree from the model’s attention patterns.
Outcome: The proposed model implicitly embeds a reasoning tree resembling the correct reasoning process within it, and detects the information from the model’s attention patterns for most examples.
Synapse: Empowering LLM Agents with Episodic-Semantic Memory via Spreading Activation (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models excel at generalized reasoning, but lack the ability to accumulate experiences and maintain narrative coherence over long horizons.
Approach: They propose a unified memory architecture that transcends static vector similarity.
Outcome: The proposed model outperforms state-of-the-art methods in temporal and multihop reasoning tasks.
FormLM: Recommending Creation Ideas for Online Forms by Modelling Semantic and Structural Information (2022.emnlp-main)

Copied to clipboard

Challenge: FormLM is a pre-trained language model for creating semi-structured forms where questions and descriptions are organized by predefined structures.
Approach: They propose to enhance pre-trained language model with form structural information to model online forms and recommend form creation ideas.
Outcome: The proposed model outperforms general-purpose language models on all tasks, with an improvement by 4.71 on Question Recommendation and 10.6 on Block Type Suggestion in terms of ROUGE-1 and Macro-F1, respectively.
SpeechLLM-as-Judges: Towards General and Interpretable Speech Quality Evaluation (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for evaluating the perceptual quality of synthetic speech are limited due to the complexity of perceptual quality factors and the diversity of speech generation tasks.
Approach: They propose a new paradigm for enabling large language models to conduct structured speech quality evaluation using a large-scale dataset.
Outcome: The proposed model performs well across tasks and languages.

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