跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2606.14990 · v3

Rational Sparse Autoencoder

首次提交 2026-06-13 06:22
版本更新 2026-11-18 09:34
雷达观测 2026-11-19 19:08
arXiv ↗PDF ↗

摘要

Sparse autoencoders (SAEs) are standard tools for mechanistic interpretability, but current SAE families are constrained by fixed encoder nonlinearities such as ReLU, JumpReLU, and TopK. This hard-codes a particular sparsity mechanism into the model and can distort the reconstruction-versus-sparsity trade-off. We introduce the Rational Sparse Autoencoder (RSAE), which replaces the fixed encoder activation with a trainable rational function. The design is motivated by a reconstruction advantage that we prove for RSAE at matched expected sparsity, with guaranteed non-inferiority to ReLU and, under stated assumptions, provable improvements over JumpReLU and TopK. The model is then implemented through a two-stage pipeline: an initialization procedure that copies the pre-trained baseline SAE weights and calibrates the scale parameters along with the rational coefficients; followed by a fine-tuning step under a sparsity-regularized reconstruction objective that holds ℓ0\ell_0 at the teacher's value. Consistent with this theory, on residual-stream activations of three open-weight language models and across all three baseline activation families, the RSAE improves reconstruction fidelity after the fine-tuning step at the same level of sparsity. These gains are consistent across host language models, across baseline activation families, and across the full range of baseline sparsity we tested, while the upgrade itself adds only 20 scalar parameters per autoencoder and one training run takes between about 4 and 25 minutes on a single consumer GPU.

作者备注:v2: M. Tang and A. Townsend added as co-authors; new theoretical analysis (Section 3, Appendix A); revised algorithm (Section 4, Appendix B.1); enhanced experiments (Section 5, Appendix B); v1 by N. Yin and Y. Yu

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v3,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v3 · 2026-11-18 09:34 · 最新观测版本

Rational Sparse Autoencoder

Sparse autoencoders (SAEs) are standard tools for mechanistic interpretability, but current SAE families are constrained by fixed encoder nonlinearities such as ReLU, JumpReLU, and TopK. This hard-codes a particular sparsity mechanism into the model and can distort the reconstruction-versus-sparsity trade-off. We introduce the Rational Sparse Autoencoder (RSAE), which replaces the fixed encoder activation with a trainable rational function. The design is motivated by a reconstruction advantage that we prove for RSAE at matched expected sparsity, with guaranteed non-inferiority to ReLU and, under stated assumptions, provable improvements over JumpReLU and TopK. The model is then implemented through a two-stage pipeline: an initialization procedure that copies the pre-trained baseline SAE weights and calibrates the scale parameters along with the rational coefficients; followed by a fine-tuning step under a sparsity-regularized reconstruction objective that holds ℓ0\ell_0 at the teacher's value. Consistent with this theory, on residual-stream activations of three open-weight language models and across all three baseline activation families, the RSAE improves reconstruction fidelity after the fine-tuning step at the same level of sparsity. These gains are consistent across host language models, across baseline activation families, and across the full range of baseline sparsity we tested, while the upgrade itself adds only 20 scalar parameters per autoencoder and one training run takes between about 4 and 25 minutes on a single consumer GPU.

查看此版本解读 →查看此版本原文 ↗