Research map / Probabilistic and scientific modeling / Molecules and proteins
Towards Understanding the Shape of Representations in Protein Language Models
Kosio Beshkov, Anders Malthe-Sorenssen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| exclude BOS/EOS token embeddings from shape-space analysis (ablation) | ESM2 and Ankh shape space analysis · SCOPe (1377 proteins) | effective dimension | Mixed |
| include BOS/EOS token embeddings (ablation) | ESM2 and Ankh context sensitivity analysis · SCOPe (1377 proteins) | graph filtration moment | No clear effect |
| increase ESM2 model size | ESM2 · SCOPe, Frechet radius | Frechet radius | No clear effect |
| increase ESM2 model size | ESM2 · SCOPe, intermediate layers | graph filtration moment (structural similarity) | Improved |
| increase ESM2 model size | ESM2 · SCOPe, last layer | graph filtration moment (structural similarity) | No clear effect |
| go deeper in layers | ESM2 · SCOPe | effective dimension of shape space | Mixed |
| use the last layer instead of the layer just before it | ESM2 · SCOPe | structural faithfulness | Hurt |
| vary interpolation spline order and number of sampled points (ablation) | SRV shape analysis · ESM2 35M | robustness of results | No clear effect |
Closest papers by meaning
- From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models (ICML 2025 · similarity 0.91)
- Elucidating the Design Space of Multimodal Protein Language Models (ICML 2025 · similarity 0.89)
- SaProt: Protein Language Modeling with Structure-aware Vocabulary (ICLR 2024 · similarity 0.89)
- FoldToken: Learning Protein Language via Vector Quantization and Beyond (AAAI 2025 · similarity 0.89)
- Protein Design with Dynamic Protein Vocabulary (NeurIPS 2025 · similarity 0.89)
- ProSST: Protein Language Modeling with Quantized Structure and Disentangled Attention (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
