EyeMulator: Improving Code Language Models by Mimicking Human Visual Attention (2026.acl-long)
Copied to clipboard
Yifan Zhang, Chen Huang, Yueke Zhang, Jiahao Zhang, Toby Jia-Jun Li, Collin McMillan, Kevin Leach, Yu Huang
| Challenge: | Code Language Models learn attention based on statistical input-output token correlations. |
| Approach: | They propose a model-agnostic technique to align CodeLLM attention with human visual attention without architectural changes. |
| Outcome: | The proposed model outperforms baselines in three languages, with gains of over 30 CodeBLEU points in translation and up to 22 BERTScore points in summarization. |
Similar Papers
From Human Reading to NLM Understanding: Evaluating the Role of Eye-Tracking Data in Encoder-Based Models (2025.acl-long)
Copied to clipboard
| Challenge: | integrating eye-tracking features into Neural Language Models does not degrade downstream task performance, enhances alignment between model attention and human attention patterns, and compresses the embedding space. |
| Approach: | They used eye-gaze data from the Ghent Eye-Tracking Corpus to investigate how integrating knowledge of human reading behavior impacts Neural Language Models. |
| Outcome: | The proposed approach does not degrade downstream task performance, enhances alignment between model attention and human attention patterns, and compresses the embedding space. |
Lending Eyesight to Language Models: Modeling and Probing Human scanpath through Transformer Decoder (2026.findings-acl)
Copied to clipboard
| Challenge: | Decoded language models have shown to exhibit striking parallels with human cognitive processes. |
| Approach: | They propose a plug-and-play module that transforms an autoregressive language model into an autorregressive eye model and probes it through a linguistic model. |
| Outcome: | The proposed module can be used to model human-like gaze shifts in language models. |
AdaV: Adaptive Text-visual Redirection for Vision-Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Vision-language models often generate excessive visual tokens, leading to poor performance . a novel training-free visual token pruning method is proposed to improve performance despite the computational cost associated with VLMs. |
| Approach: | They propose a training-free visual token pruning method that reduces biased token pruning . they plan to open-source the code upon publication . |
| Outcome: | The proposed method reduces biased token pruning and enhances model robustness with limited visual token budget. |
Don’t Miss the Forest for the Trees: Attentional Vision Calibration for Large Vision Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Vision Language Models suffer from hallucinations, attributing incorrect or misleading features to images. |
| Approach: | They propose a test-time approach that recalibrates the influence of blind tokens . they identify blind token by analyzing layer-wise attention distributions over image tokens. |
| Outcome: | The proposed approach reduces hallucinations in large vision language models . it uses a contrastive decoding strategy to balance the influence of blind tokens . |
HiPrune: Hierarchical Attention for Efficient Token Pruning in Vision-Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for visual token pruning lack insight into the intrinsic property of the vision encoder . et al., 2017: 99.3% of task accuracy with only 1/3 of the tokens. |
| Approach: | They propose a model-agnostic token pruning method that trains without training . they propose 'HiPrune' method which prunes visual tokens according to their attention . |
| Outcome: | The proposed method achieves 99.3% of task accuracy with only 1/3 of the tokens . it reduces inference FLOPs by 58.7% and maintains 99.99% accuracy with 2/9 tokens. |
CodeT5+: Open Code Large Language Models for Code Understanding and Generation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing code LLMs adopt a specific architecture or rely on a unified encoder-decoder network for downstream tasks, lacking flexibility to operate in the optimal architecture for a particular task. |
| Approach: | They propose to initialize code LLMs with frozen off-the-shelf LLM and explore instruction-tuning to align with natural language instructions. |
| Outcome: | The proposed model outperforms open-source LLMs on 20 code-related benchmarks. |
Mixture of Decoding: An Attention-Inspired Adaptive Decoding Strategy to Mitigate Hallucinations in Large Vision-Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Vision-Language Models (LVLMs) have impressive capabilities across visual tasks, yet they remain hindered by the persistent challenge of hallucinations. |
| Approach: | They propose a novel approach that dynamically adapts decoding strategies by evaluating the correctness of the model’s attention on image tokens to distinguish the correct attention. |
| Outcome: | Extensive experiments show that the proposed approach outperforms existing decoding methods across multiple mainstream benchmarks, effectively mitigating hallucinations in LVLMs. |
Beyond Label Attention: Transparency in Language Models for Automated Medical Coding via Dictionary Learning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Current efforts in interpretability of medical coding rely heavily on label attention mechanisms, which often leads to the highlighting of extraneous tokens irrelevant to the ICD code. |
| Approach: | They propose to leverage dictionary learning to extract sparsely activated representations from dense language models embedded in superposition to facilitate accurate interpretability. |
| Outcome: | The proposed model extracts sparsely activated representations from dense language models in superposition, even when the highlighted tokens are medically irrelevant. |
Target Foresight Based Attention for Neural Machine Translation (N18-1)
Copied to clipboard
| Challenge: | Empirical experiments on Chinese-to-English and Japanese-to English datasets show that the proposed attention model delivers significant improvements in terms of alignment error rate and BLEU. |
| Approach: | They propose to explicitly access the target foresight word in the attention model to improve alignment and translation accuracy. |
| Outcome: | Empirical results show that the proposed model improves alignment error rate and BLEU on Chinese-to-English and Japanese-toEnglish datasets. |
Recurrent Attention for Neural Machine Translation (2021.emnlp-main)
Copied to clipboard
| Challenge: | Recent research questions the importance of dot-product self-attention in Transformer models and shows that most attention heads learn simple positional patterns. |
| Approach: | They propose a novel mechanism to replace dot-product self-attention with a recurrent atteNtion mechanism that directly learns attention weights without token-to-token interaction. |
| Outcome: | The proposed model outperforms the Transformer model on translation tasks with fewer parameters and inference time. |