paper-with-me

Papers

Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects

2026-05-30 · Hwiyeong Lee, Ingyu Bang, Uiji Hwang, Hyelim Lim, Taeuk Kim arxiv

While sparse autoencoders provide features more interpretable than individual neurons, reliably characterizing them remains challenging. We propose Query Lens, which extends Logit Lens to enable more comprehensive and faithful interpretations of sparse features. By jointly considering encoder-side key features and decoder-side value features, we identify both the inputs that activate a feature and the outputs it promotes. We also account for indirect, module-mediated effects that arise when the feature is processed by downstream modules, going beyond the direct effect captured by Logit Lens. In experiments, we find that Query Lens yields coherent token signatures for features that remain uninterpretable under Logit Lens. Finally, we propose the Subspace Channel Hypothesis, suggesting that downstream modules read features through layer-specific subspaces.

📄 PDF Abstract BibTeX arXiv:2606.07617

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Neural Response Interpretation through the Lens of Critical Pathways

2021-03-31 · CVPR 2021 1 · Ashkan Khakzar, Soroosh Baselizadeh, Saurabh Khanduja, Christian Rupprecht 외

Is critical input information encoded in specific sparse pathways within the neural network? In this work, we discuss the problem of identifying these critical pathways and subsequently leverage them for interpreting the…

Sparse Readout Prism: Explaining Logit-Lens Scores in Features Instead of Tokens

2026-09-01 · Matteo He, William F. Shen, Xinchi Qiu, Nicholas D. Lane hf

A language model's prediction of its next token develops across layers, and lens methods track this process by decoding intermediate hidden states into tokens. But a lens reading reflects both the hidden state and the re…

Explain like I am BM25: Interpreting a Dense Model's Ranked-List with a Sparse Approximation

2023-04-25 · Michael Llordes, Debasis Ganguly, Sumit Bhatia, Chirag Agarwal

Neural retrieval models (NRMs) have been shown to outperform their statistical counterparts owing to their ability to capture semantic meaning via dense document representations. These models, however, suffer from poor i…

Retrieval

Interpreting Attention Layer Outputs with Sparse Autoencoders

2024-06-25 · Connor Kissane, Robert Krzyzanowski, Joseph Isaac Bloom, Arthur Conmy 외

Decomposing model activations into interpretable components is a key open problem in mechanistic interpretability. Sparse autoencoders (SAEs) are a popular method for decomposing the internal activations of trained trans…

SAFER: Probing Safety in Reward Models with Sparse Autoencoder

2025-07-01 · Wei Shi, Ziyuan Xie, Sihang Li, Xiang Wang arxiv

Reinforcement learning from human feedback (RLHF) is a key paradigm for aligning large language models (LLMs) with human values, yet the reward models at its core remain largely opaque. In this work, we present Sparse Au…

Reinforcement Learning