Papers by Antoine Simoulin
Contrasting distinct structured views to learn sentence embeddings (2021.eacl-srw)
Copied to clipboard
| Challenge: | Existing methods to build sentence embeddings rely on a similar Recurrent Neural Network (RNN) heterogeneity of performances across models and tasks makes us assume some structures might be better adapted given the considered task or sentence. |
| Approach: | They propose a self-supervised method that builds sentence embeddings from syntactic structures . they hypothesize that some linguistic representations might be better adapted given the task . |
| Outcome: | The proposed method outperforms comparable methods on several tasks from standard sentence embedding benchmarks. |
DocAgent: A Multi-Agent System for Automated Code Documentation Generation (2025.acl-demo)
Copied to clipboard
| Challenge: | Existing methods for generating documentation using Large Language Models (LLMs) produce incomplete, unhelpful, or factually incorrect outputs. |
| Approach: | They propose a novel collaborative system that uses topological code processing for incremental context building to generate documentation by agents. |
| Outcome: | The proposed system outperforms baselines in completeness, helpfulness, and truthfulness evaluations. |
Unifying Parsing and Tree-Structured Models for Generating Sentence Semantic Representations (2022.naacl-srw)
Copied to clipboard
| Challenge: | Existing tree-based models require handannotated data to be trained. |
| Approach: | They propose a tree-based model that learns its composition function together with its structure. |
| Outcome: | The proposed model outperforms existing models on downstream tasks and is competitive with Bert base model. |
Memory-Efficient Fine-Tuning of Transformers via Token Selection (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for fine-tuning require caching of intermediate activations to update weights during the backward pass. |
| Approach: | They develop a method to reduce memory usage in fine-tuning of transformers by backpropagating through just a subset of input tokens. |
| Outcome: | The proposed method reduces memory usage and memory footprint on large transformer models . it can be easily combined with existing methods like LoRA, reducing memory cost . |
Code to Think, Think to Code: A Survey on Code-Enhanced Reasoning and Reasoning-Driven Code Intelligence in LLMs (2025.emnlp-main)
Copied to clipboard
Dayu Yang, Tianyang Liu, Daoan Zhang, Antoine Simoulin, Xiaoyi Liu, Yuwei Cao, Zhaopu Teng, Xin Qian, Grey Yang, Jiebo Luo, Julian McAuley
| Challenge: | Recent breakthrough models like OpenAI-o1 and DeepSeek-R1 show powerful task-solving capabilities, particularly advances in reasoning. |
| Approach: | They propose future research directions that may deepen the synergy, ultimately advancing LLM performance in both complex reasoning and code intelligence. |
| Outcome: | The proposed research may deepen the synergy, ultimately advancing LLM performance in both complex reasoning and code intelligence. |
How Many Layers and Why? An Analysis of the Model Depth in Transformers (2021.acl-srw)
Copied to clipboard
| Challenge: | In deep transformers, weights are tied across layers, resulting in multiple layers. |
| Approach: | They propose a variant of Albert that adapts the number of layers for each token of the input. |
| Outcome: | The proposed model implements the key specificity of Albert and iterates on token representations over time. |