Sparse autoencoders decompose Earth embeddings into sparse features
On AlphaEarth, trained SAEs leave eight times less variance unexplained than a frozen random decoder.
Stories per day · 30 days
1 RSS
Relatedinterpretability8mechanistic-interpretability3ai-safety1ai-text-detection1alphaearth1behavioral-modes1classifiers1compositionality1concept-erasure1data-efficiency1dictionary-learning1diffusion-models1earth-observation1embeddings1feature-reproducibility1foundation-models1gemma1gemma-21
16 stories, newest first · one card per story
Stories tagged #sparse-autoencoders
User embeddings are superpositions of interpretable behavioral modes
Sparse autoencoders recover named behavioral modes from embeddings and support zero-shot natural-language queries.
ConceptACP erases diffusion concepts with sparse autoencoders
On SDXL-Turbo one-step generation, forgetting success rises 36.34 points over the best baseline.
Sparse autoencoders turn neural AI-text detectors into readable classifiers
Classifiers using only interpretable SAE features beat neural and stylometric detectors on out-of-distribution text.
ConSDL improves sparse dictionary consistency over SAE baselines
On 16 synthetic settings recovery was near perfect; consistency rose 17.6% for dictionaries and 35.9% for coefficients.
Is AI interpretability close to a breakthrough? A new essay has researchers divided
Nick Cammarata argues famous polysemantic neurons were misread, claiming most Qwen3-32B neurons have one job.
7
Redistribution score tracks recurring SAE feature shifts
On two frozen safety readouts, median supports of about 20 features reached failure AUROCs of 0.935 and 0.958.
Behavioral Compositionality of Sparse Autoencoder Representations
SAE features are more compositional than random bases, but faithful explanations still need pairwise interactions.
Diminishing Returns of Data in Sparse Autoencoders: Functional Behavior does not Guarantee Feature Reproducibility
SAE reconstruction converges on far less data, but one-to-one feature matches keep rising with more tokens.
SAE-IDF matches symbols across systems using sparse autoencoder codes
On 12 tasks across six symbolic systems, SAE-IDF beats the same LLM's dense baseline by up to 45 points.
SOUL unlearning reduces state hallucination in robot VLA policies
The method forgets hallucination-linked sparse features while retaining successful manipulation behavior in simulation and real robots.
Parts-of-Speech as Emergent Categories in SAE Latent Space
Sparse autoencoder latents encode part-of-speech in distributed groups, not one-to-one grammatical features.
Local Sparsity Enables Unsupervised LLM Safety Detection
Proposes locally masked sparse-autoencoder anomaly detection that flags unsafe LLM outputs unsupervised, using only 1-2% of SAE neurons.
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
SAEScientist-Bench tests whether AI agents can autonomously run SAE interpretability research on Gemma-2-9B-IT; frontier agents trail expert baselines.
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
SAEScientist-Bench evaluates whether AI agents can autonomously conduct SAE interpretability research in Gemma-2-9B-IT, finding frontier agents trail expert baselines.
LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders
Researchers use sparse autoencoders to localize trigger-based backdoor mechanisms in 1B and 8B LLMs, finding detection features differ from causal control features.