RESEARCH RADAR · DAILY
July 24, 2026
0 peer-reviewed · 2 preprints · 0 forum/blog
Mech Interp · AI Security · Text Diffusion LMs
Also today — items 4–10
07
mech-interp

Statistically Grounded Sparse-Feature Interventions for Activation-Space Control in Large Language Models

Six-condition reliability filter + Borda consensus ranking (F-test, KSG MI, Cohen's d) selects SAE features for steering without a learned objective. Key finding: substantial gap between activation shift and generation quality preservation.

08
mech-interp AI security

Sparse Autoencoders are Capable LLM Jailbreak Mitigators

CC-Delta (Context-Conditioned Delta Steering) identifies jailbreak-relevant SAE features by contrasting activations with/without jailbreak context, then steers in SAE latent space. Outperforms dense steering on all 4 tested models across 12 jailbreak types.

8 entries removed on 2026-09-10 as repeats of earlier reports: 2505.19223 (first covered 2026-07-15), 2607.15893 (first covered 2026-07-21), 2606.04027 (first covered 2026-07-02), 2606.20560 (first covered 2026-07-01), 2602.05859 (first covered 2026-07-01), 2607.07903 (first covered 2026-07-12), 2606.00023 (first covered 2026-07-10), 2607.01774 (first covered 2026-07-16).

← all Research Radar issues · gussand · source