Pith. sign in

REVIEW 12 cited by

SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.09532 v4 pith:XNQCZQMD submitted 2025-03-12 cs.LG cs.CL

Adam Karvonen , Can Rager , Johnny Lin , Curt Tigges , Joseph Bloom , David Chanin , Yeu-Tong Lau , Eoin Farrell
show 7 more authors
Callum McDougall Kola Ayonrinde Demian Till Matthew Wearden Arthur Conmy Samuel Marks Neel Nanda
This is my paper
classification cs.LGcs.CL
keywords metricssaesacrossarchitecturespracticalproxysaebenchautoencoders
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Sparse autoencoders (SAEs) are a popular technique for interpreting language model activations, and there is extensive recent work on improving SAE effectiveness. However, most prior work evaluates progress using unsupervised proxy metrics with unclear practical relevance. We introduce SAEBench, a comprehensive evaluation suite that measures SAE performance across eight diverse metrics, spanning interpretability, feature disentanglement and practical applications like unlearning. To enable systematic comparison, we open-source a suite of over 200 SAEs across eight recently proposed SAE architectures and training algorithms. Our evaluation reveals that gains on proxy metrics do not reliably translate to better practical performance. For instance, while Matryoshka SAEs slightly underperform on existing proxy metrics, they substantially outperform other architectures on feature disentanglement metrics; moreover, this advantage grows with SAE scale. By providing a standardized framework for measuring progress in SAE development, SAEBench enables researchers to study scaling trends and make nuanced comparisons between different SAE architectures and training methodologies. Our interactive interface enables researchers to flexibly visualize relationships between metrics across hundreds of open-source SAEs at: www.neuronpedia.org/sae-bench

Discussion (0). Sign in to comment.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations

    cs.LG 2026-05 conditional novelty 7.0 of 10

    SA-GSAE with Bi-Jump-ReLU enables one latent to encode both polarities of anticorrelated features, Pareto-dominating or matching full-width gated SAEs while reducing dead latents by up to 500x on some LLM hookpoints.

  2. Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures

    cs.CL 2025-06 conditional novelty 7.0 of 10

    A two-level mixture-of-experts sparse autoencoder models parent and child concepts together, improving reconstruction and reducing feature redundancy on Gemma 2-2B activations compared to flat top-k SAEs.

  3. ECG-InterpBench: Benchmarking the Interpretability of ECG Foundation Models with Matched-Scale Sparse Autoencoders

    cs.LG 2026-07 conditional novelty 6.0 of 10

    With matched-scale sparse autoencoders, HuBERT-ECG best preserves its ECG representation while ECG-JEPA best exposes clinical measurements through single features — a leader split that repeats on MIMIC-IV-ECG.

  4. Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects

    cs.LG 2026-07 conditional novelty 6.0 of 10

    A single-token feature's causal necessity under zero-ablation depends on which SAE family found it: GemmaScope and BatchTopK features stay causally anchored while LlamaScope features are locally redundant.

  5. Decoder-Preserving Sparse Autoencoders: Which Readouts Survive Sparse Compression?

    cs.LG 2026-07 accept novelty 6.0 of 10

    A new SAE objective penalizes disagreement between ridge prediction operators, preserving more linear readouts at equal reconstruction error.

  6. Persistent Sparse Autoencoders: Learning Feature Timescales in Language Models

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Persistent SAEs learn per-feature persistence coefficients from reconstruction, splitting features into fast local detectors and slow topic-tracking states that retain prompt-injection signals over long contexts.

  7. Resa: Transparent Reasoning Models via SAEs

    cs.CL 2025-06 conditional novelty 6.0 of 10

    SAE-Tuning, a sparse-autoencoder-guided SFT procedure, elicits RL-comparable reasoning in 1.5B models from CoT-free QA data at about $1 and 20 minutes of training.

  8. Stable and Steerable Sparse Autoencoders with Weight Regularization

    stat.ML 2026-03 conditional novelty 5.0 of 10

    L2 weight regularization in TopK SAEs increases cross-seed feature overlap and roughly doubles measured steering success on Pythia-70M, at the cost of collapsing most latents to zero.

  9. Distribution-Aware Feature Selection for SAEs

    cs.LG 2025-08 conditional novelty 5.0 of 10

    Sampled-SAE pre-selects a candidate pool of features using batch-level norms or entropy before batch top-K, creating a tunable family that trades reconstruction fidelity for improved probing and reduced absorption on ...

  10. On the transferability of Sparse Autoencoders for interpreting compressed models

    cs.LG 2025-07 conditional novelty 5.0 of 10

    Pruning a pretrained sparse autoencoder can produce an interpretability tool for a WANDA-pruned LLM that is roughly comparable to retraining an SAE on the pruned model, though with notable caveats in the reported metrics.

  11. Evaluating SAE interpretability without explanations

    cs.LG 2025-07 conditional novelty 5.0 of 10

    SAE latent interpretability can be scored directly from activation examples via intruder detection and embedding clustering, with LLM scores correlating strongly with human scores.

  12. Train One Sparse Autoencoder Across Multiple Sparsity Budgets to Preserve Interpretability and Accuracy

    cs.LG 2025-05 conditional novelty 5.0 of 10

    HierarchicalTopK trains a single sparse autoencoder that reconstructs transformer activations well at many sparsity levels, matching or beating separate per-level models.

Pith tools