AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

ProtSAE: Disentangling and Interpreting Protein Language Models via Semantically-Guided Sparse Autoencoders

Xiangyu Liu, Haodi Lei, Yi Liu, Yang Liu, Wei Hu

AAAI 2026 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 3 citations (OpenAlex)

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 changedWhereMeasuredEffect
detach W_pred when tying W_def so reconstruction loss does not update it (add detach) (ablation)ProtSAE · BPO dataset, varying L0AUCImproved
add axiom loss from ELEmbeddings domain knowledge (ablation)ProtSAE · BPO dataset, varying L0AUC and Loss RecoveredImproved
add semantic bias z_bias (forced activation) (ablation)ProtSAE · BPO dataset, varying L0proportion of predicted activations used in decodingImproved
add learnable rescaling r_pred for defined weights (ablation)ProtSAE · BPO dataset, varying L0AUC and Loss RecoveredImproved
add semantic guidance during SAE training (ProtSAE vs TopK SAE)TopK SAE · probing-based interpretation, avg of three datasetsAUC: .565 -> .797Improved
add semantic guidance during SAE training (ProtSAE vs TopK SAE)TopK SAE · probing-based interpretation, avg of three datasetsFmax: .444 -> .579Improved
use ProtSAE features instead of linear probe on PLM hidden statesESM2-15B probing · probing-based interpretation, avg of three datasetsAUPR: AUPR .522 -> .487; Fmax .537 -> .579Mixed
add semantic guidance during SAE training (ProtSAE vs baseline SAEs)SAE reconstruction · BPO dataset, varying sparsityLoss RecoveredNo clear effect
activate ProtSAE concept features during generation (steering)ESM2-15B masked reconstruction · seven concepts, 50% masked tokensTM-score and RMSDImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy