Papers by Dong Fang

30 papers
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.
CART: A Generative Cross-Modal Retrieval Framework With Coarse-To-Fine Semantic Modeling (2025.acl-long)

Copied to clipboard

Challenge: Cross-modal retrieval tasks are used to retrieve data from one modality or another based on a query from another modality.
Approach: They propose a generative cross-modal retrieval framework based on coarse-to-fine semantic modeling . they propose combining K-Means and RQ-VAE to discretize multimodal data into token sequences that support autoregressive generation.
Outcome: The proposed framework achieves excellent performance and efficiency in multimodal retrieval tasks.
Bridging the Gap between Synthetic and Authentic Images for Multimodal Machine Translation (2023.emnlp-main)

Copied to clipboard

Challenge: Existing models require associated image with input sentence, which is difficult to satisfy at inference.
Approach: They propose to use synthetic and authentic images to generate translations using text-to-image generation models.
Outcome: The proposed model achieves state-of-the-art performance on En-De and En-Fr datasets while remaining independent of authentic images during inference.
IntentCoding: Amplifying User Intent in Code Generation (2026.findings-acl)

Copied to clipboard

Challenge: IntentCoding captures the influence of user intent by masking out the intent, and integrates seamlessly with existing decoding procedures.
Approach: They propose a decoding strategy that captures the influence of user intent by masking out the intent and applies a multi-strength ensemble mechanism to amplify the effect of user intention during generation.
Outcome: The proposed model significantly improves both constraint satisfaction and functional correctness compared to greedy decoding approaches.
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.
DevEval: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories (2024.findings-acl)

Copied to clipboard

Challenge: Existing benchmarks are poorly aligned with real-world code repositories and are insufficient to evaluate the coding abilities of Large Language Models (LLMs).
Approach: They propose a repository-level benchmark named DevEval to evaluate LLMs' coding abilities in real-world code repositories.
Outcome: The proposed benchmarks show that the LLMs perform better in real-world code repositories than existing benchmarks.
LearnAlign: Data Selection for LLM Reinforcement Learning with Improved Gradient Alignment (2026.findings-acl)

Copied to clipboard

Challenge: Reinforcement learning with verifiable rewards (RLVR) is a key technique for enhancing LLMs’ reasoning abilities, yet its data inefficiency remains a major bottleneck.
Approach: They propose a gradient-alignment-based method which intelligently selects the learnable and representative training reasoning data for RLVR post-training.
Outcome: Experiments on five reasoning benchmarks show that the proposed method significantly reduces training data requirements while improving performance.
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.
OpenWebVoyager: Building Multimodal Web Agents via Iterative Real-World Exploration, Feedback and Optimization (2025.acl-long)

Copied to clipboard

Challenge: Existing studies focus on building text-only agents in synthetic environments where the reward signals are clearly defined.
Approach: They propose a multimodal web agent that can autonomously conduct real-world exploration and improve itself after each iteration.
Outcome: The proposed agent improves itself after each iteration, demonstrating strong performance across multiple test sets.
WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models (2026.acl-long)

Copied to clipboard

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.
AnchorAttention: Difference-Aware Sparse Attention with Stripe Granularity (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for large language models with extended context lengths face significant computational challenges during the prefill phase.
Approach: They propose a difference-aware, dynamic sparse attention mechanism that efficiently identifies critical attention regions at a finer stripe granularity while adapting to global contextual information.
Outcome: The proposed model achieves a speedup of 1.44 while maintaining higher recall rates.
Assessing Combinational Generalization of Language Models in Biased Scenarios (2022.aacl-short)

Copied to clipboard

Challenge: Existing work focuses on assessing in-domain knowledge, but shedding light on what pre-trained Language Models learn is important.
Approach: They propose a method to assess a PLM's generalization capacity in biased scenarios by combining component combinations where it could be easy for the PLMs to learn shortcuts from the training corpus.
Outcome: The proposed model can overcome distribution shifts in the training corpus and with sufficient data.
KELE: A Multi-Agent Framework for Structured Socratic Teaching with Large Language Models (2025.findings-emnlp)

Copied to clipboard

Challenge: Socratic teaching places high demands on teachers’ expertise and real-time feedback capabilities, making it difficult to scale in large educational settings.
Approach: They propose a multi-agent framework for structured Socratic teaching with LLMs that integrates a structured SocRule and a consultant-teacher collaborative teaching mechanism.
Outcome: The proposed framework outperforms existing LLMs in natural language generation and dialogue comprehension in the classroom.
AutoBreach: Universal and Adaptive Jailbreaking with Efficient Wordplay-Guided Optimization via Multi-LLMs (2025.findings-naacl)

Copied to clipboard

Challenge: Existing jailbreak research exhibits limitations in universality, validity, and efficiency . Existing methods for jailbreaking LLMs have limited validity and effectiveness .
Approach: They propose a black-box approach that uses wordplay-guided mapping rule sampling to create universal adversarial prompts.
Outcome: The proposed method efficiently identifies security vulnerabilities across various LLMs, achieving an average success rate of over 80% with fewer than 10 queries.
MedDialog: Large-scale Medical Dialogue Datasets (2020.emnlp-main)

Copied to clipboard

Challenge: telemedicine is a medical practice that provides patient care remotely using video conferencing tools.
Approach: They build large-scale medical dialogue datasets to facilitate research . they pretrain several models on the Chinese MedDialog dataset and compare their performance .
Outcome: The proposed datasets show that models trained on MedDialog can generate doctor-like medical dialogues.
Knowledge-to-Verification: Exploring RLVR for LLMs in Knowledge-Intensive Domains (2026.acl-long)

Copied to clipboard

Challenge: Recent large language models (LLMs) have demonstrated remarkable progress in reasoning, but their applications on knowledge-intensive domains have not been explored due to the scarcity of high-quality verifiable data.
Approach: They propose a framework that extends reinforcement learning with verifiable rewards (RLVR) to knowledge-intensive domains through automated verififiability data synthesis while enabling verification of the LLM's reasoning process.
Outcome: Extensive experiments show that the proposed framework enhances the reasoning of large language models in knowledge-intensive domains without significantly compromising the model’s general capabilities.
DivScene: Towards Open-Vocabulary Object Navigation with Large Vision Language Models in Diverse Scenes (2025.findings-emnlp)

Copied to clipboard

Challenge: Large Vision-Language Models (LVLMs) have achieved significant progress in tasks like visual question answering and document understanding.
Approach: They introduce DivScene, a large-scale dataset with 4,614 houses across 81 scene types and 5,707 kinds of target objects.
Outcome: The proposed dataset provides a much greater diversity of target objects and scene types than existing datasets, enabling a comprehensive task evaluation.
WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback (2025.findings-emnlp)

Copied to clipboard

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.
Enhancing Multimodal Unified Representations for Cross Modal Generalization (2025.findings-acl)

Copied to clipboard

Challenge: Existing studies on discrete unified representations overlook important distinctions between different dimensions of features.
Approach: They propose to use a codebook to optimize unified representations from pretraining and fine- and coarse-grained disentangling to optimize the representations.
Outcome: The proposed methods improve the interpretability of multimodal unified representations . they use training-free optimization of codebook and fine and coarse cross-modal disentangling .
Implicit Deep Latent Variable Models for Text Generation (D19-1)

Copied to clipboard

Challenge: Variational auto-encoders have been used for text generation but their representation power is limited due to two reasons.
Approach: They advocate sample-based representations of variational distributions for natural language . they further develop an LVM to directly match the aggregated posterior to the prior .
Outcome: The proposed model can be viewed as a natural extension of VAEs with a regularization of maximizing mutual information, mitigating the "posterior collapse" issue.
Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to overcome object hallucination are limited . Existing mitigations include costly retraining and a training-free inference framework .
Approach: They propose a training-free inference framework that simulates a metacognitive self-correction process.
Outcome: The proposed framework reduces object hallucination rates by 12.67% on MMHal-Bench and improves accuracy by 5.8% on POPE.
Attention Entropy is a Key Factor: An Analysis of Parallel Context Encoding with Full-attention-based Pre-trained Language Models (2025.acl-long)

Copied to clipboard

Challenge: Large language models have demonstrated remarkable performance across a wide range of language tasks due to their remarkable ability in context modeling.
Approach: They propose to use parallel context encoding to reduce attention entropy by incorporating attention sinks and selective mechanisms to reduce irregular attention . they also propose to incorporate attention sink mechanisms into the parallel encoded context to reduce the irregular attention.
Outcome: The proposed methods lower irregular attention entropy and narrow performance gaps.
CAFE: Retrieval Head-based Coarse-to-Fine Information Seeking to Enhance Multi-Document QA Capability (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods to extend context length of Large Language Models (LLMs) still struggle with retrieval and reasoning in long context inputs.
Approach: They propose a coarse-to-fine method to enhance multi-document question-answering capacities by removing background and distracting documents.
Outcome: Experiments show that CAFE outperforms baseline methods on multiple documents.
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.
Train in Vain: Functionality-Preserving Poisoning to Prevent Unauthorized Use of Code Datasets (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for dataset poisoning require full-dataset poison, which breaks code compilability.
Approach: They propose a functionality-preserving poisoning approach that injects short, compilable weak-use fragments into executed code paths.
Outcome: The proposed method contaminates 10% of the dataset while maintaining 100% compilability and functional correctness.
Towards Robust Temporal Activity Localization Learning with Noisy Labels (2024.lrec-main)

Copied to clipboard

Challenge: Existing methods for temporal activity localization are expensive and difficult to satisfy due to subjective labeling.
Approach: They propose a new TAL setting where a TAL model should be robust to mixed training data with noisy moment boundaries.
Outcome: The proposed method is significantly more robust to noisy training data than existing methods.
Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification (2026.findings-acl)

Copied to clipboard

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.
Assessing the Reliability of Word Embedding Gender Bias Measures (2021.emnlp-main)

Copied to clipboard

Challenge: Various measures have been proposed to quantify human-like social biases in word embeddings, but they can suffer from measurement error.
Approach: They propose to assess the reliability of word embedding gender bias measures by examining their reliability across different choices of random seeds, scoring rules and words.
Outcome: The proposed measures can suffer from measurement error, and the results inform better design of word embedding gender bias measures.
Continual Named Entity Recognition without Catastrophic Forgetting (2023.emnlp-main)

Copied to clipboard

Challenge: Named Entity Recognition (CNER) is a burgeoning area of research . a new paradigm has ushered NER into a non-entity type at the current step t .
Approach: They propose a pooled feature distillation loss that skillfully navigates the trade-off between retaining knowledge of old entity types and acquiring new ones.
Outcome: The proposed method outperforms state-of-the-art approaches on ten CNER settings using three datasets.

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