Replicating an interpretability paper with Opus 5.5: 20 of 22 results confirmed on Qwen3-32B
Sauers_ · x · 2026-10-17
The author used Opus 5.5 to reproduce experiments from "Neural Networks Are Already Clean", writing their own code to validate 20 of 22 published results on Qwen3-32B. The one miss stems from direct silencing vs. letting the network compensate, per Opus, which also flagged a second "unplottable" deviation. Overall a solid replication—and a working demo of LLM-driven interpretability reproduction.
More from Models
- Kaggle launches a 'pelican riding a bike' arena: GPT-6 Astra leads, Gemini 4 Argon close behind — MeganRisdal · 2026-10-17
- Dev Slams Corporate HR-Style AI: I Want a Model That Tells Me When I'm an Idiot — VraserX · 2026-10-17
- Perplexity's Decider V1.2 decision models land on Vercel AI Gateway — AravSrinivas · 2026-10-17
- Grok Bot update: 3x faster computer tasks, X Chat access, new connectors and plugin marketplace — XFreeze · 2026-10-17
- Neel Nanda: no alignment faking found in Gemini 4 Argon RL training audit of 10K rollouts — NeelNanda5 · 2026-10-17
- Chinese Frontier Models Go Quiet: Kimi K3 at 92 Days, K3.1 and GLM-5.5 Rumored Soon — haider1 · 2026-10-17