Update README.md

nulone · web-flow · commit f880fde2ae02 · 2026-02-26T21:40:16.000+07:00
diff --git a/README.md b/README.md
@@ -1,3 +1,5 @@
+[![Verify Numbers](https://github.com/nulone/sae-consciousness-steering-pitfalls/actions/workflows/verify.yml/badge.svg)](https://github.com/nulone/sae-consciousness-steering-pitfalls/actions/workflows/verify.yml)
+
 # SAE Consciousness Steering: A Multi-Model Null Result
 
 I tried to use contrastive SAE discovery to find features that control how language models answer consciousness-related questions. After 9 experiments on Gemma 3 4B and Gemma 3 12B (plus a qualitative Neuronpedia label search on Llama 3.3 70B), I found no evidence of causal consciousness features with this pipeline. The contrastive method finds punctuation, Japanese grammar, and self-referential discourse markers — not consciousness.

Original file line number	Diff line number	Diff line change
`@@ -1,3 +1,5 @@`
	`1`	`+[![Verify Numbers](https://github.com/nulone/sae-consciousness-steering-pitfalls/actions/workflows/verify.yml/badge.svg)](https://github.com/nulone/sae-consciousness-steering-pitfalls/actions/workflows/verify.yml)`
	`2`	`+`
`1`	`3`	`# SAE Consciousness Steering: A Multi-Model Null Result`
`2`	`4`
`3`	`5`	`I tried to use contrastive SAE discovery to find features that control how language models answer consciousness-related questions. After 9 experiments on Gemma 3 4B and Gemma 3 12B (plus a qualitative Neuronpedia label search on Llama 3.3 70B), I found no evidence of causal consciousness features with this pipeline. The contrastive method finds punctuation, Japanese grammar, and self-referential discourse markers — not consciousness.`