Papers by Lei Ji

33 papers
Too Long, Do Re-weighting for Efficient LLM Reasoning Compression (2026.acl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have recently achieved remarkable progress on complex reasoning tasks by leveraging extended Chain-of-Thought (CoT) techniques.
Approach: They propose a method that uses Extended Chain-of-Thought (EFT) to reduce the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning.
Outcome: The proposed method reduces the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning.
ParaMac: A General Unsupervised Paraphrase Generation Framework Leveraging Semantic Constraints and Diversifying Mechanisms (2022.findings-emnlp)

Copied to clipboard

Challenge: Existing unsupervised methods for paraphrase generation are weak in semantic equivalence or expression diversity.
Approach: They propose a framework for unsupervised paraphrase generation that employs multi-aspect equivalence constraints and multi-granularity diversifying mechanisms to achieve good semantic equvalence and expressive diversity.
Outcome: The proposed framework achieves 9.1% and 3.3% absolute gains over previous SOTA on Quora and MSCOCO and can improve to 18.0% and 4.6% on GLUE.
RotateCT: Knowledge Graph Embedding by Rotation and Coordinate Transformation in Complex Space (2022.coling-1)

Copied to clipboard

Challenge: Existing knowledge graph embedding methods fail to model non-commutative composition patterns . Existing methods are limited to complex space, resulting in a large number of parameters.
Approach: They propose a knowledge graph embedding method that transforms the coordinates of each entity and then represents each relation as a rotation from head entity to tail entity in complex space.
Outcome: The proposed method outperforms state-of-the-art methods on link prediction and path query answering.
DynamixSFT: Dynamic Mixture Optimization of Instruction Tuning Collections (2026.findings-acl)

Copied to clipboard

Challenge: Several studies rely on additional models to optimize mixtures.
Approach: They propose a method that dynamically optimizes instruction-tuning dataset mixtures by prior-scaled Boltzmann Exploration and a multi-armed bandit setup.
Outcome: The proposed method improves the TÜLU-2-mixture and TÜLO-3-mixtures across 10 benchmarks while introducing minimal computational overhead over naive sampling.
Hashing based Efficient Inference for Image-Text Matching (2021.findings-acl)

Copied to clipboard

Challenge: Recent work on image-text matching has focused on exploring interactions between images and sentences to improve performance without considering inference efficiency.
Approach: They propose a hashing-based efficient inference module which can be plugged into existing frameworks to speed up inference step without reducing retrieval performance.
Outcome: The proposed module can be plugged into existing framework to speed up inference step without reducing retrieval performance.
AesX: Enhance Your Images with Stunning Aesthetic Beauty (2026.acl-industry)

Copied to clipboard

Challenge: Existing models do not analyze human preferences at a finer granularity, which leads to quality issues.
Approach: They propose a set of preference indicators across two major dimensions, text-image consistency and aesthetic quality, and a generative framework to steer the model toward a generation path that more closely aligns with human aesthetic sensibilities.
Outcome: The proposed model improves target recognition accuracy and overall visual aesthetic presentation by focusing on human preferences.
Preserving Knowledge Invariance: Rethinking Robustness Evaluation of Open Information Extraction (2023.emnlp-main)

Copied to clipboard

Challenge: Existing evaluation benchmarks focus on pairwise matching, ignoring robustness . current models exhibit frustrating degradation, with a maximum drop of 23.43 F1 score .
Approach: They propose a benchmark that simulates the evaluation of open information extraction models in the real world . they perform experiments on typical models published in the last decade and a representative large language model .
Outcome: The proposed model is rated robust on a knowledge-invariant clique with different syntactic and expressive forms.
Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for reweighting data mixtures rely on manual designation with certain heuristics based on intuition or empirical results.
Approach: They propose a model-based framework that learns to re-weight domains by reinforcement learning on large quantities of data mixing trajectories with corresponding feedback from an evaluation environment.
Outcome: The proposed framework outperforms baselines in achieving balanced performance across source and target fields and domain spaces without retraining.
GEM: A General Evaluation Benchmark for Multimodal Tasks (2021.findings-acl)

Copied to clipboard

Challenge: Existing datasets that focus on natural language tasks are not considered as a general evaluation benchmark for multimodal tasks.
Approach: They present a general evaluation benchmark for multimodal tasks, GEM 1 . they compare it with existing multimodal vision-language datasets .
Outcome: The proposed model is compared with existing vision-language datasets focusing on natural language tasks . it is the largest vision-linguistic dataset covering image-language tasks and video-language task at the same time .
CSMCIR: CoT-Enhanced Symmetric Alignment with Memory Bank for Composed Image Retrieval (2026.findings-acl)

Copied to clipboard

Challenge: Existing approaches to search for images using single-modality are limited by representation space fragmentation.
Approach: They propose a unified representation framework that achieves efficient query-target alignment . they introduce a multi-level Chain-of-Thought prompting strategy that guides MLMs to generate discriminative, semantically compatible captions for target images .
Outcome: The proposed framework achieves efficient query-target alignment through synergistic components.
Explore the Reasoning Capability of LLMs in the Chess Testbed (2025.naacl-short)

Copied to clipboard

Challenge: a recent study shows that large language models struggle with long-term, complex reasoning tasks.
Approach: They propose to integrate annotated strategy and tactic into large language models to improve reasoning capability.
Outcome: The proposed model performs better than GPT, Claude, and Gemini models . it integrates annotated strategy and tactic into the model .
Neural Machine Translation for Agglutinative Languages via Data Rejuvenation (2025.acl-srw)

Copied to clipboard

Challenge: Recent years, advances in Neural Machine Translation (NMT) heavily rely on large-scale parallel corpora.
Approach: They propose to combine fine-grained inactive sample identification with target-side rejuvenation to improve translation quality from agglutinative languages.
Outcome: The proposed framework improves on four low-resource agglutinative language tasks.
Simple and Effective Graph-to-Graph Annotation Conversion (2022.coling-1)

Copied to clipboard

Challenge: Existing work on graph-structured annotation conversions has focused on feature-based models which are not easily applicable to new conversions.
Approach: They propose two graph-to-graph conversion approaches which use pseudo data and inherit parameters to guide conversions respectively.
Outcome: The proposed approaches outperform strong baselines with higher conversion score on a graph-structured dataset and other datasets.
Hierarchical Context-aware Network for Dense Video Event Captioning (2021.acl-long)

Copied to clipboard

Challenge: Existing technologies for dense video event captioning generate fine-grained captions for all events in a long untrimmed video.
Approach: They propose a hierarchical context-aware network for dense video event captioning to capture context from various aspects.
Outcome: The proposed model outperforms the existing model on youcook2 and activitynet . it generates coherent captions for events in a long untrimmed video .
CONE: An Efficient COarse-to-fiNE Alignment Framework for Long Video Temporal Grounding (2023.acl-long)

Copied to clipboard

Challenge: Existing work on video temporal grounding for long videos is limited by existing datasets.
Approach: They propose a query-guided window selection strategy and a coarse-to-fine mechanism to speed up inference for long videos.
Outcome: The proposed framework accelerates inference time by 2x on Ego4D-NLQ and 15x on MAD while keeping SOTA results.
Control Image Captioning Spatially and Temporally (2021.acl-long)

Copied to clipboard

Challenge: Existing methods to generate image captions with user intention are still under exploration.
Approach: They propose a model that connects Contrastive constraints and Attention Guidance in a loop manner and engages explicit spatial and temporal constraints to the generating process.
Outcome: The proposed model improves performance on a trace-controlled image captioning task.
Resource-Friendly Dynamic Enhancement Chain for Multi-Hop Question Answering (2025.findings-acl)

Copied to clipboard

Challenge: Existing approaches to solve multi-hop question answering challenges require multiple rounds of retrieval and iterative generation.
Approach: They propose a framework that decomposes complex questions into coherent subquestions . it then iteratively refines these subquests through context-aware rewriting to generate effective query formulations.
Outcome: The proposed framework performs on par with or surpasses state-of-the-art benchmarks while significantly reducing token consumption.
LLM-Based Multi-Agent Systems are Scalable Graph Generative Models (2025.findings-acl)

Copied to clipboard

Challenge: Social graphs are mathematical structures stem from pairwise interactions between entities through nodes and edges.
Approach: They propose a framework for dynamic, text-attributed social graph generation that simulates the temporal node and edge generation processes for zero-shot social graphs.
Outcome: The proposed framework improves macroscopic graph structure metrics by 11% . the proposed model can generate graphs with up to 100,000 nodes or 10 million edges .
T2A-Feedback: Improving Basic Capabilities of Text-to-Audio Generation via Fine-grained AI Feedback (2025.acl-long)

Copied to clipboard

Challenge: Text-to-audio (T2A) models still struggle to satisfy human preferences for prompt-following and acoustic quality when generating complex multi-event audio.
Approach: They propose to use AI feedback learning to enhance basic capabilities of text-to-audio models . they use a large audio preference dataset to evaluate the model's capabilities .
Outcome: The proposed model improves in simple and complex scenarios with AI feedback learning.
GRACE: Gradient Harmonized and Cascaded Labeling for Aspect-based Sentiment Analysis (2020.findings-emnlp)

Copied to clipboard

Challenge: Existing studies ignore aspect terms interaction when labeling polarities . aspect terms extraction and aspect sentiment classification are two fundamental tasks .
Approach: They propose a GRadient hArmonized and CascadEd labeling model to solve the imbalance issue . they extend the gradient harmonized mechanism used in object detection to aspect-based sentiment analysis .
Outcome: The proposed model achieves consistency improvement on multiple benchmark datasets and generates state-of-the-art results.
Generative Prompt Internalization (2025.naacl-long)

Copied to clipboard

Challenge: Prompts used in large language model based applications are often fixed and lengthy, leading to significant computational overhead.
Approach: They propose a method that internalizes complex prompts using a joint training approach and a data synthesis technique that auto-collects conversational datasets by swapping roles of agent and environment.
Outcome: The proposed method internalizes complex prompts across agent-based applications and generates the content along with reasons for why it should change accordingly.
LayerNorm Induces Recency Bias in Transformer Decoders (2026.findings-acl)

Copied to clipboard

Challenge: Existing studies show that stacking causal self-attention layers alone induces a positional bias in attention scores toward earlier tokens, but this differs from the bias toward later tokens observed in Transformer decoders, known as recency bias.
Approach: They propose to stack causal self-attention layers and layer norm to induce recency bias in Transformer decoders by analyzing the interaction between causal self and other architectural components.
Outcome: The proposed method provides new theoretical insights into how positional information interacts with architectural components and suggests improvements in positional encoding strategies.
Dense Procedure Captioning in Narrated Instructional Videos (P19-1)

Copied to clipboard

Challenge: Existing models for video dense captioning learn video segments and generate captions without considering transcripts.
Approach: They propose a model to generate procedure captions from narrated instructional videos . they extract procedures by a cross-modality module and generate captions by encoding video frames and transcripts within each extracted procedure.
Outcome: The proposed model can extract procedures from narrated instructional videos and generate procedure captions by encoding video frames and transcripts.
Syntactically Robust Training on Partially-Observed Data for Open Information Extraction (2022.findings-emnlp)

Copied to clipboard

Challenge: Open Information Extraction models have shown promising results with sufficient supervision, but the syntactic distribution of training data is partially observable in comparison to the real world.
Approach: They propose a syntactically robust training framework that enables models to be trained on a multi-paraphrase distribution based on diverse paraphrase generation.
Outcome: The proposed framework can be applied to other syntactic partial observable domains.
MM-MATH: Advancing Multimodal Math Evaluation with Process Evaluation and Fine-grained Classification (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing benchmarks for multimodal reasoning in large multimodal models are underperforming on multimodal tasks.
Approach: They propose a benchmark for multimodal reasoning in large multimodal models, MM-MATH . MM's process evaluation employs LMM-as-a-judge to automatically analyze solution steps . diagram misinterpretation is the most common error, they find .
Outcome: The proposed model achieves only 31% accuracy, compared to 82% for humans.
LongAlign: A Recipe for Long Context Alignment of Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing studies to build long context language models focus on context extension and continual training on long text.
Approach: They propose a recipe for instruction fine-tuning on input sequences of similar length . they adopt packing and sorted batching strategies to speed up supervised fine-uning .
Outcome: The proposed model outperforms existing recipes for LLMs in long context tasks by 30% while maintaining proficiency in handling short, generic tasks.
ABC-Bench: Benchmarking Agentic Backend Coding in Real-World Development (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models (LLMs) have redefined the role of AI in software engineering . current benchmarks focus on localized code generation, but neglect dynamic, full-process requirements of real-world engineering.
Approach: They propose a benchmark to evaluate agentic backend coding within a realistic, executable workflow.
Outcome: The ABC-Bench benchmark evaluates agentic backend coding within a realistic, executable workflow.
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.
Learning Shared Semantic Space for Speech-to-Text Translation (2021.findings-acl)

Copied to clipboard

Challenge: End-to-end speech translation (ST) has been treated as an independent task . however, the modality gap has rendered MT data and its end-to end models incompatible with their ST counterparts.
Approach: They propose to bridge the representation gap between text and audio inputs by projecting audio and text features to a common semantic representation.
Outcome: The proposed model improves performance on two ST benchmarks and achieves 27.1 BLEU on MuST-C EN-DE.
Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations (2025.emnlp-main)

Copied to clipboard

Challenge: LLMs often use assertive language when making false claims, resulting in harm and loss of trust.
Approach: They find that a mismatch between semantic and verbal uncertainty is a better predictor of hallucinations than semantic uncertainty alone.
Outcome: a new study shows that mismatch between semantic and verbal uncertainty is better predictor of hallucinations than semantic uncertainty alone.
Refining and Synthesis: A Simple yet Effective Data Augmentation Framework for Cross-Domain Aspect-based Sentiment Analysis (2024.findings-acl)

Copied to clipboard

Challenge: Aspect-based Sentiment Analysis (ABSA) data augmentation has attracted increasing attention in recent years due to data sparsity.
Approach: They propose a framework to augment ABSA data using pseudo labels for target domain . they refine generated labeled data using a natural language inference filter .
Outcome: The proposed framework outperforms 7 strong baselines on 4 kinds of ABSA tasks.
What Factors Influence LLMs’ Judgments? A Case Study on Question Answering (2024.lrec-main)

Copied to clipboard

Challenge: Existing studies indicate that Large Language Models perform at a level comparable to humans with advantages of speed and cost-effectiveness in different fields.
Approach: They propose to introduce four unexplored factors and a new dimension of question difficulty to provide a more comprehensive understanding of LLMs’ judgments across varying question intricacies.
Outcome: The proposed dimensions of question difficulty and answer quantity provide valuable insights into optimizing LLMs’ performance as judges.

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