Papers by Haitao Mi
Verified Critical Step Optimization for LLM Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Critical Step Optimization (CSO) focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Approach: | They propose a method which focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Outcome: | The proposed method outperforms the existing methods on GAIA-Text-103 and XBench-DeepSearch while requiring supervision at only 16% of trajectory steps. |
A Dialogue-based Information Extraction System for Medical Insurance Assessment (2021.findings-acl)
Copied to clipboard
Shuang Peng, Mengdi Zhou, Minghui Yang, Haitao Mi, Shaosheng Cao, Zujie Wen, Teng Xu, Hongbin Wang, Lei Liu
| Challenge: | a new system that integrates advanced NLP technologies for medical insurance assessment is proposed . the average time cost of the procedure is reduced from 55 minutes to 35 minutes . |
| Approach: | They propose a dialogue-based information extraction system that integrates advanced NLP technologies for medical insurance assessment. |
| Outcome: | The proposed system reduces the time cost of the procedure from 55 minutes to 35 minutes and saves 30% human resources cost compared with the previous offline procedure. |
Friend-training: Learning from Models of Different but Related Tasks (2023.eacl-main)
Copied to clipboard
| Challenge: | Current self-training methods focus on improving model performance on a single task. |
| Approach: | They propose a cross-task self-training framework where models trained to do different tasks are used in iterative training, pseudo-labeling, and retraining processes to help each other for better selection of pseudo-labeled labels. |
| Outcome: | The proposed framework achieves the best performance compared to baselines on two dialogue understanding tasks. |
Recall with Reasoning: Chain-of-Thought Distillation for Mamba’s Long-Context Memory and Extrapolation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing long-context memory methods such as Mamba struggle with long-constituency when the length of the processed text exceeds the model's training length. |
| Approach: | They propose a method that uses chain-of-thought summarization to teach Mamba to actively recall and reason over long contexts. |
| Outcome: | Experiments on LONGMEMEVAL and HELMET show that RwR outperforms existing long-term memory methods while preserving short-context capabilities. |
Self-Consistency Boosts Calibration for Math Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing solutions for math reasoning tasks use semantic parsing or AST decoding, but performance can degrade dramatically even with slight changes to the questions. |
| Approach: | They propose three calibration methods based on self-consistency for math reasoning tasks. |
| Outcome: | The proposed methods bridge model confidence and accuracy better than existing methods based on p(True) or logit. |
Inconsistent dialogue responses and how to recover from them (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing methods to assess and bolster utterance consistency of chat systems have been shown difficult to detect. |
| Approach: | They propose to use annotators to write dialogue responses and recovery utterances to assess and bolster utteration consistency of chat systems. |
| Outcome: | The proposed dataset significantly improves the detection and resolution of inconsistencies in chat conversations. |
Self-Alignment for Factuality: Mitigating Hallucinations in LLMs via Self-Evaluation (2024.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to addressing factual inaccuracies require high-quality human factuality annotations to mitigate these hallucinations. |
| Approach: | They propose to leverage the self-evaluation capability of an LLM to provide training signals that steer the model towards factuality. |
| Outcome: | The proposed approach significantly improves factual accuracy over LLMs across three key knowledge-intensive tasks on TruthfulQA and BioGEN. |
A Knowledge Plug-and-Play Test Bed for Open-domain Dialogue Generation (2024.lrec-main)
Copied to clipboard
| Challenge: | Knowledge-based open-domain dialogue generation aims to build chit-chat systems that talk to humans using mined support knowledge. |
| Approach: | They propose a benchmark for evaluating multi-source dialogue knowledge selection and response generation using Wikipedia's wizard of Wikipedia. |
| Outcome: | The proposed benchmark is called multi-source Wizard of Wikipedia (Ms.WoW) it contains clean support knowledge, grounded at the utterance level and partitioned into multiple knowledge sources. |
WebRollback: Enhancing Web Agents with Explicit Rollback Mechanisms (2026.eacl-short)
Copied to clipboard
| Challenge: | Recent studies have adopted a greedy one-way search strategy to deal with dynamic web environments. |
| Approach: | They propose to integrate a rollback mechanism into web agents to allow them to revert back to a previous state in navigation trajectory. |
| Outcome: | The proposed method is able to revert back to a previous state in its navigation trajectory, allowing the models to directly control the search process. |
R2D2: Recursive Transformer based on Differentiable Tree for Interpretable Hierarchical Language Modeling (2021.acl-long)
Copied to clipboard
| Challenge: | Existing models with stacked layers do not explicitly model hierarchical structure of language understanding. |
| Approach: | They propose a recursive Transformer model based on differentiable CKY style binary trees to emulate hierarchical composition process. |
| Outcome: | The proposed model can predict words given their left and right abstraction nodes. |
Cross-lingual Text-to-SQL Semantic Parsing with Representation Mixup (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Experimental results show that Rex can benefit from cross-lingual training and improve the effectiveness of semantic parsers. |
| Approach: | They propose a Representation Mixup Framework for effectively exploiting translations in the cross-lingual Text-to-SQL task. |
| Outcome: | The proposed framework can benefit from cross-lingual training and improve the effectiveness of semantic parsers, achieving state-of-the-art performance. |
Crossing the Reward Bridge: Expanding Reinforcement Learning with Verifiable Rewards Across Diverse Domains (2026.acl-long)
Copied to clipboard
| Challenge: | Reinforcement learning with verifiable rewards (RLVR) has been effective on structured tasks, but its reliance on simple, rule-based verifiers creates a bottleneck. |
| Approach: | They propose a framework that uses a generative verifier to provide soft, probabilistic rewards. |
| Outcome: | The proposed framework outperforms existing models up to 10x their size and can be scalable and effective. |
Entropy Guided Extrapolative Decoding to Improve Factuality in Large Language Models (2025.coling-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) exhibit impressive natural language capabilities but suffer from hallucination – generating content that does not align with realworld facts. |
| Approach: | They propose to extrapolate critical token probabilities beyond the last layer to improve decoding by manipulating the predicted distributions at inference time. |
| Outcome: | The proposed methods surpass state-of-the-art on multiple datasets by large margins. |
Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data (2026.acl-short)
Copied to clipboard
| Challenge: | Strong base models saturate benchmarks, resulting in weaker performance, a paradox . a new approach to Reinforcement Learning (RL) is needed to improve performance . |
| Approach: | They propose a method that uses constrained uniform top-k sampling to flatten the local optimization landscape by sampling uniformly from constrained high-confidence candidates. |
| Outcome: | Experiments show that the proposed approach prevents policy degeneration and boosts out-of-domain generalization. |
Low-Bit Quantization Favors Undertrained LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Larger models or those trained on fewer tokens exhibit less quantization-induced degradation (QiD), while smaller, well-trained models face significant performance losses. |
| Approach: | They propose to use QiD to measure an LLM’s training levels and determine the number of training tokens required for fully training LLMs of various sizes. |
| Outcome: | The proposed scaling laws can predict the quantization performance of different-sized LLMs trained with tokens. |
WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models (2026.acl-long)
Copied to clipboard
Rui Wang, Ce Zhang, Jun-Yu Ma, Jianshu Zhang, Hongru Wang, Yi Chen, Boyang Xue, Tianqing Fang, Zhisong Zhang, Hongming Zhang, Haitao Mi, Dong Yu, Kam-Fai Wong
| Challenge: | Existing agentic systems are retrieval-heavy but reasoning-light . current systems lack compositional reasoning, a key component of deep research . |
| Approach: | They propose a data synthesis pipeline WebAggregator to shift agentic paradigm . they use Proactive Explorer to collect interconnected knowledge and Compositional Logic Proposer to weave knowledge into complex questions . |
| Outcome: | The proposed pipeline surpasses GPT-4.1 and matches Claude-3.7-Sonnet on GAIA, WebWalkerQA, and XBench. |
OpenFact: Factuality Enhanced Open Knowledge Extraction (2023.tacl-1)
Copied to clipboard
Linfeng Song, Ante Wang, Xiaoman Pan, Hongming Zhang, Dian Yu, Lifeng Jin, Haitao Mi, Jinsong Su, Yue Zhang, Dong Yu
| Challenge: | Existing OIE systems organize knowledge into subject-relation-object (SRO) triplets, and they use templates to extract such knowledge triplet. |
| Approach: | They propose a framework to handle expressiveness and groundedness in OpenFact . they propose to use templates, extra constraints, and adopt human efforts to ensure that most triplets contain enough details. |
| Outcome: | The proposed framework improves expressiveness and groundedness of OpenFact . it is more accurate and denser than OPIEC-Linked, which is grounded to Wikidata . |
SafeConv: Explaining and Correcting Conversational Unsafe Behavior (2023.acl-long)
Copied to clipboard
| Challenge: | Existing datasets do not provide enough annotation to explain unsafe behavior . current chatbots generate toxic and offensive responses, which can be dangerous . |
| Approach: | They construct a dataset called SafeConv that provides comprehensive annotations for chatbots . they compare safe alternatives to rewrite unsafe responses . |
| Outcome: | The proposed model can explain unsafe behavior and detoxify chatbots, the authors show . the proposed model is able to detect unsafe utterances, extract unsafe spans, and convert unsafe responses to safe versions. |
Your Reasoning Model is Secretly a Reward Model - Optimization-Free Verification from Experience (2026.acl-long)
Copied to clipboard
| Challenge: | Existing verifiers operate on the surface text or on confidence proxies derived from token probabilities, which can be brittle. |
| Approach: | They propose a training-free, non-parametric verifier that summarizes each reasoning trace by an activation delta and compares it to two class centroids computed from labeled experience. |
| Outcome: | The proposed model improves selection and reranking on large and less-calibrated models. |
Improving LLM Generations via Fine-Grained Self-Endorsement (2024.findings-acl)
Copied to clipboard
| Challenge: | Recent large language models (LLMs) have demonstrated remarkable capabilities but can still fail frequently on knowledge-intensive tasks. |
| Approach: | They propose a self-endorsement framework that leverages fine-grained fact-level comparisons across multiple sampled responses. |
| Outcome: | The proposed framework can improve factuality of generations with simple prompts across scales of LLMs. |
Don’t Get Lost in the Trees: Streamlining LLM Reasoning by Overcoming Tree Search Exploration Pitfalls (2025.acl-long)
Copied to clipboard
Ante Wang, Linfeng Song, Ye Tian, Dian Yu, Haitao Mi, Xiangyu Duan, Zhaopeng Tu, Jinsong Su, Dong Yu
| Challenge: | Recent advances in tree search algorithms guided by verifiers have significantly enhanced the reasoning capabilities of large language models (LLMs), but at the cost of increased computational resources. |
| Approach: | They propose an e ffici ent tree sear ch framework that is a plug-and-play system compatible with various tree search algorithms. |
| Outcome: | The proposed framework reduces computational costs and prioritizes resource allocation to harder tasks (Levels 3-4) over simpler ones (Level 1-2), addressing both over-exploration in basic problems and under-exploation in complex cases. |
Save the Good Prefix: Precise Error Penalization via Process-Supervised RL to Enhance LLM Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing reinforcement learning methods rely on sparse outcome rewards, which fail to credit correct intermediate steps in partially successful solutions. |
| Approach: | They propose a process reward model that rewards correct steps only when they detect errors . they propose VPPO, which rewards the correct prefix and an erroneous suffix . |
| Outcome: | a new approach outperforms sparse-reward RL and prior PRM-guided baselines on Pass@1 and Pass@K . a process reward model (PRM) outperformed sparser-rebound RL on multiple reasoning benchmarks . |
Self-Tuning: Instructing LLMs to Effectively Acquire New Knowledge through Self-Teaching (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to keeping large language models current involve continued pre-training on new documents. |
| Approach: | They propose a learning framework that augments documents with knowledge-intensive tasks created in a self-supervised manner, focusing on memorization, comprehension, and self-reflection. |
| Outcome: | The proposed learning framework improves an LLM’s ability to acquire new knowledge from unseen raw documents through self-teaching. |
More Than Spoken Words: Nonverbal Message Extraction and Generation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies focus on extracting NMs from small-scale well-structured corpora such as movie scripts wherein NM is enclosed in parentheses by scriptwriters, which greatly decreases the difficulty of extraction. |
| Approach: | They propose to extract nonverbal messages (NMs) from written text and NMs from spoken text by using a semi-supervised learning algorithm. |
| Outcome: | The extracted NMs can generate more relevant, valid, and factually consistent NM than the purely supervised generator. |
WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback (2025.findings-emnlp)
Copied to clipboard
Minda Hu, Tianqing Fang, Jianshu Zhang, Jun-Yu Ma, Zhisong Zhang, Jingyan Zhou, Hongming Zhang, Haitao Mi, Dong Yu, Irwin King
| Challenge: | Web agents powered by Large Language Models lack the ability to perform in uncertain web environments. |
| Approach: | They propose to reconstruct web agents' reasoning skills into chain-of-thought rationales by fine-tuning their LLM backbone into a web-based model. |
| Outcome: | The proposed approach significantly improves the agent self-improving benchmark OpenWebVoyager, demonstrating that it can be used to improve the agent's reasoning skills. |
WebEvolver: Enhancing Web Agent Self-Improvement with Co-evolving World Model (2025.emnlp-main)
Copied to clipboard
| Challenge: | Agent self-improvement, where agents train their underlying Large Language Model (LLM) on self-sampled trajectories, shows promising results but often stagnates in web environments due to limited exploration and under-utilization of pretrained web knowledge. |
| Approach: | They propose a co-evolving Large Language Model (LLM) that predicts the next observation based on current observation and action within the web environment. |
| Outcome: | The proposed framework shows that agents can perform better in real-world web environments without using any distillation from more powerful close-sourced models. |
Learning a Grammar Inducer from Massive Uncurated Instructional Videos (2022.emnlp-main)
Copied to clipboard
| Challenge: | aims to find more accurate syntactic grammars for accompanying text using video data. |
| Approach: | They build a video-aided grammar induction model that can learn video-span correlation without manual features. |
| Outcome: | The proposed model can learn video-span correlation without manual features adopted by previous systems. |
Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding (2026.acl-long)
Copied to clipboard
| Challenge: | Graphical User Interface (GUI) grounding requires mapping natural language instructions to precise pixel coordinates due to visually homogeneous elements and dense layouts. |
| Approach: | They propose to replace static consistency strategies with a learnable selection mechanism that selects the optimal target by critiquing its own proposals rendered on the screenshot. |
| Outcome: | The proposed model significantly improves both grounding and critiquing capabilities over 6 benchmarks. |
EconProver: Towards More Economical Test-Time Scaling for Automated Theorem Proving (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have recently advanced the field of Automated Theorem Proving (ATP) Existing cost analyses regulate only the number of sampling passes, ignoring the substantial disparities in sampling costs. |
| Approach: | They propose to integrate two complementary methods into a unified EconRL pipeline to increase pass rates under constrained sampling passes. |
| Outcome: | The proposed method reduces token usage and sample passes while maintaining the original performance. |
Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification (2026.findings-acl)
Copied to clipboard
Yuxuan Wan, Tianqing Fang, Zaitang LI, Yintong Huo, Wenxuan Wang, Haitao Mi, Dong Yu, Michael R. Lyu
| Challenge: | Recent advances in Deep Research Agents (DRAs) are transforming automated knowledge discovery and problem-solving. |
| Approach: | They propose an inference-time scaling of verification wherein an agent self-improves at test time by evaluating its generated answers. |
| Outcome: | The proposed model outperforms vanilla agent-as-judge and LLM judge baselines by 12%–48% in meta-evaluation F1 score. |
Fast-R2D2: A Pretrained Recursive Neural Network based on Pruned CKY for Grammar Induction and Text Representation (2022.emnlp-main)
Copied to clipboard
| Challenge: | Chart-based models have shown great potential in unsupervised grammar induction, running recursively and hierarchically, but requiring O(n3) time-complexity. |
| Approach: | They propose a model-guided pruning method that scales to large language model pretraining by introducing a heuristic pruning method. |
| Outcome: | The proposed method significantly improves grammar induction quality and achieves competitive results in downstream tasks. |
Bi-level Finetuning with Task-dependent Similarity Structure for Low-resource Training (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to fine tune a large language model in low-resource settings are limited in their expressiveness or rely on task-independent knowledge. |
| Approach: | They propose a framework where all parameters are finetuned with task-dependent information from the training data only. |
| Outcome: | The proposed framework outperforms baseline models on several classification datasets in low-resource scenarios. |