Papers by Heng Zhou

28 papers
From Lists to Emojis: How Format Bias Affects Model Alignment (2025.acl-long)

Copied to clipboard

Challenge: Format biases in reinforcement learning from human feedback are underexplored . despite its effectiveness, RLHF faces challenges, including policy and regulatory constraints .
Approach: They extend the study of preference biases beyond verbosity bias to a wider range of format biase . they show that with a small amount of biased data, they can inject significant bias into the reward model .
Outcome: The proposed approach can be easily exploited by large language models to achieve higher rankings on popular benchmarks like AlpacaEval and LMSYS Chatbot Arena.
KIA: Knowledge-Guided Implicit Vision-Language Alignment for Chest X-Ray Report Generation (2025.coling-main)

Copied to clipboard

Challenge: Existing reports on medical images and reports lack fine-grained cross-modal interaction, leading to insufficient understanding of detailed information.
Approach: They propose a framework for establishing cross-modal semantic alignment in radiology report pairs using knowledge-guided implicit vision-language alignment.
Outcome: KIA improves understanding of medical images and reports by incorporating medical knowledge to enhance pathological observation and anatomical landm.
Reaction Miner: An Integrated System for Chemical Reaction Extraction from Textual Data (2023.emnlp-demo)

Copied to clipboard

Challenge: Reaction Miner is a system designed to extract chemical reactions from raw scientific PDFs.
Approach: They propose a system that extracts chemical reactions directly from raw scientific PDFs.
Outcome: The proposed system can extract chemical reactions from raw scientific PDFs.
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.
Human-in-the-loop Schema Induction (2023.acl-demo)

Copied to clipboard

Challenge: Existing approaches to event-centric natural language understanding (NLU) have been limited to linear and temporal ones.
Approach: They propose a human-in-the-loop schema induction system powered by GPT-3 . they show that it transfers to new domains more easily than previous approaches .
Outcome: The proposed system transfers to new domains more easily than previous approaches and reduces human curation.
Video-Text Prompting for Weakly Supervised Spatio-Temporal Video Grounding (2024.emnlp-main)

Copied to clipboard

Challenge: Existing methods extract each candidate tube feature independently by cropping objects from video frame feature, discarding all contextual information such as position change and inter-entity relationship.
Approach: They propose to use video-text prompts to construct candidate feature instead of cropping tube region from feature map . they also propose negative contrastive samples whose candidate object is erased instead of being highlighted .
Outcome: The proposed methods surpass existing weakly-supervised methods by a great margin . they draw visual markers over objects tubes as video prompts .
BadWindtunnel: Defending Backdoor in High-noise Simulated Training with Confidence Variance (2025.findings-acl)

Copied to clipboard

Challenge: Current backdoor attack defenders in NLP typically involve data reduction or model pruning, risking losing crucial information.
Approach: They propose a backdoor defender that allows precise control over training conditions to model backdoor learning behavior without affecting the final model.
Outcome: The proposed model reduces the backdoor learning behavior without affecting the final model.
From Implicit Graph Encoding to Explicit Evidence: A Training-Free LLM Framework for Temporal Knowledge Graph Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Existing Large Language Models (LLMs) struggle with implicit modality alignment and suboptimal graph linearization.
Approach: They propose a training-free, test-time adaptive framework that reframes TKG prediction as explicit evidence-driven reasoning.
Outcome: ExE-LLM outperforms fully trained graph neural networks on four benchmarks . it achieves SOTA performance in inductive settings, significantly outperforming fully trained neural networks .
M2PO: Multi-Perspective Multi-Pair Preference Optimization for Machine Translation (2026.acl-long)

Copied to clipboard

Challenge: prevailing methods for machine translation are often hindered by misleading reward signals.
Approach: They propose a framework that aligns large language models to human preferences . they propose 'M2PO' to correct the bias towards partial errors .
Outcome: The proposed framework outperforms open-source models and achieves parity with proprietary models.
FlashAudio: Rectified Flow for Fast and High-Fidelity Text-to-Audio Generation (2025.acl-long)

Copied to clipboard

Challenge: Recent advances in latent diffusion models (LDMs) have markedly enhanced text-to-audio generation, yet their iterative sampling processes impose substantial computational demands, limiting practical deployment.
Approach: They propose to learn straight flow for fast simulation by using flashAudio with rectified flows and immiscible flow to minimize the total distance of data-noise pairs in a batch vias assignment.
Outcome: The proposed method can learn straight flow for fast simulations and reduce noise distribution.
Document-level Event Factuality Identification via Machine Reading Comprehension Frameworks with Transfer Learning (2022.coling-1)

Copied to clipboard

Challenge: Document-level Event Factuality Identification (DEFI) is a fundamental and crucial task in NLP.
Approach: They propose a framework for document-level event factuality identification (DEFI) they propose to use Span-Extraction and Multiple-Choice to model DEFI as machine reading comprehension tasks .
Outcome: The proposed model outperforms state-of-the-art models on a document-based event factuality task . it uses Span-Extraction (Ext) and Multiple-Choice (Mch) knowledge to extract knowledge from large-scale MRC corpus .
RESIN: A Dockerized Schema-Guided Cross-document Cross-lingual Cross-media Information Extraction and Event Tracking System (2021.naacl-demos)

Copied to clipboard

Challenge: We present a new information extraction system that can construct temporal event graphs from news documents.
Approach: They propose a temporal event graph extraction system that can extract news documents . they extend the system from sentence-level event extraction to cross-document cross-media event extraction .
Outcome: The proposed system can extract temporal event graphs from news documents in multiple languages and multiple data modalities.
Specialization without Sparsity: Efficient and Expressive Split-Path Experts for LLM Fine-Tuning (2026.findings-acl)

Copied to clipboard

Challenge: Parameter-efficient fine-tuning (PEFT) is a low-cost alternative to full fine-timing due to the massive overhead.
Approach: They propose a Mixture-of-Experts approach that enhances specialization while maintaining low resource overhead.
Outcome: The proposed approach outperforms or matches state-of-the-art methods on GLUE, GSM8K, MBPP, and a text rewriting task from SmolTalk.
Paper Abstract Writing through Editing Mechanism (P18-2)

Copied to clipboard

Challenge: a paper abstract writing system can automatically generate an abstract from a title . a typical recurrent neural network (RNN) based approach easily loses focus.
Approach: They propose a paper abstract writing system that automatically generates an abstract from a title.
Outcome: The proposed system passes Turing tests by junior domain experts and non-experts at a rate up to 80%.
ReSo: A Reward-driven Self-organizing LLM-based Multi-Agent System for Reasoning Tasks (2025.emnlp-main)

Copied to clipboard

Challenge: Multi-agent systems (MAS) are limited by poor flexibility and scalability, with underdeveloped optimization strategies.
Approach: They propose a task graph generation and a reward-driven two-stage agent selection process to integrate multi-agent systems to improve their reasoning capabilities.
Outcome: The proposed model outperforms existing methods on Math-MAS and SciBench-MAS SciBech, while other methods completely fail.
Text2DB: Integration-Aware Information Extraction with Large Language Model Agents (2024.findings-acl)

Copied to clipboard

Challenge: Current methods for information extraction (IE) focus on integrating IE output with the database . a long-overlooked question is what counts as "relevant knowledge"
Approach: They propose a task that emphasizes integration of IE output and the database . they introduce a benchmark and an LLM agent framework for this task .
Outcome: The proposed task integrates IE output and the target database (or knowledge base) it meets common demands such as data infilling, row population, and column addition .
New Frontiers of Information Extraction (2022.naacl-tutorials)

Copied to clipboard

Challenge: Information extraction (IE) is the process of automatically extracting structural information from unstructured or semi-structured data.
Approach: This tutorial will provide an introduction to recent advances in IE by answering several important research questions.
Outcome: The tutorial will address several important research questions and outline directions for further investigation.
Event Causality Identification via Derivative Prompt Joint Learning (2022.coling-1)

Copied to clipboard

Challenge: Existing methods for event causality identification lack annotated data, and they lack the ability to identify explicit and implicit causality.
Approach: They propose a derivative prompt joint learning model which leverages potential causal knowledge in the pre-trained language model to tackle the data scarcity problem.
Outcome: The proposed model can identify explicit and implicit causality on two benchmark datasets and it has great advantages over previous methods.
Symbolization, Prompt, and Classification: A Framework for Implicit Speaker Identification in Novels (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing methods for speaker identification in novel dialogues are limited to handling explicit narrative patterns and complex cases.
Approach: They propose a framework which identifies implicit speakers in novels via symbolization, prompt, and classification.
Outcome: The proposed framework outperforms existing methods by 4.8% accuracy on the web novel collection, which reduces 47% of speaker identification errors, and outperfies the emerging ChatGPT.
ERICA: Improving Entity and Relation Understanding for Pre-trained Language Models via Contrastive Learning (2021.acl-long)

Copied to clipboard

Challenge: Existing pre-training objectives do not explicitly model relational facts in text . Experimental results show that ERICA can improve typical PLMs on several language understanding tasks, including relation extraction, entity typing and question answering.
Approach: They propose a contrastive learning framework ERICA to obtain a deep understanding of entities and relations in text.
Outcome: The proposed framework can improve PLMs on several language understanding tasks, especially under low-resource settings.
Fine-grained Coordinated Cross-lingual Text Stream Alignment for Endless Language Knowledge Acquisition (D18-1)

Copied to clipboard

Challenge: a novel information network decipherment paradigm is proposed for fine-grained coordinated cross-lingual text stream alignment.
Approach: They propose to use Burst Information Networks as media to represent text streams . they propose a simple yet effective information network decipherment algorithm with diverse clues .
Outcome: The proposed approach outperforms existing approaches on bilingual lexicon extraction from coordinated text streams and can harvest high-quality alignments from large amounts of streaming data for endless language knowledge mining.
Incentivizing Parametric Knowledge via Reinforcement Learning with Verifiable Rewards for Cross-Cultural Entity Translation (2026.acl-long)

Copied to clipboard

Challenge: Current systems often fall short of this goal in settings where translation hinges on culturally grounded entities such as books, films, places, songs and idioms.
Approach: They propose a framework that anchors supervision on a verifiable, entity-level reward signal and incorporates lightweight structural gates to stabilize optimization.
Outcome: The proposed framework improves on XC-Translate and shows that it can learn a robust reasoning process rather than imitating reference translations.
SARA: Unlocking Multilingual Knowledge in Mixture-of-Experts via Semantically Anchored Routing Alignment (2026.findings-acl)

Copied to clipboard

Challenge: Low-resource language tokens are often routed to different experts than those activated by high-resourced inputs, which hinders their efficacy in multilingual contexts.
Approach: They propose a framework to transfer specialized capabilities from high-resource languages as anchors to low-resourced languages by using a symmetric Jensen-Shannon constraint.
Outcome: The proposed framework outperforms standard instruction tuning on 5 low-resource languages and 3 benchmarks.
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.
Non-Sequential Graph Script Induction via Multimedia Grounding (2023.acl-long)

Copied to clipboard

Challenge: Existing scripts for everyday tasks are presented in a linear manner, which does not reflect the flexibility displayed by people executing tasks in real life.
Approach: They propose to use loosely aligned videos to train a non-sequential graph script induction task by using a multimodal framework to ground procedural videos to WikiHow textual steps.
Outcome: The proposed model outperforms the WikiHow linear baseline by 48.76% . it can predict future steps given a partial step sequence and generate explicit graph scripts .
Do LVLMs Understand Charts? Analyzing and Correcting Factual Errors in Chart Captioning (2024.findings-acl)

Copied to clipboard

Challenge: LVLMs are known for producing text that is factually inconsistent with visual input . factuality of generated captions for structured visuals has not been studied as much .
Approach: They propose a typology of factual errors in captions generated by large vision-language models . they propose CHOCOLATE, a visual entailment model that outperforms current models based on this analysis .
Outcome: The proposed model outperforms current models in evaluating caption factuality.
Enhanced Chart Understanding via Visual Language Pre-training on Plot Table Pairs (2023.findings-acl)

Copied to clipboard

Challenge: Existing methods to understand chart plots are difficult to apply to visual-language tasks.
Approach: They propose a V+L model that learns how to interpret table information from chart images via cross-modal pre-training on plot table pairs.
Outcome: The proposed model outperforms state-of-the-art models on the chartQA benchmark by over 8% performance gains.

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