dataset

active

dataset:146-self-correction-episodes-from-llama-3-3-70b

146 Self-Correction Episodes from Llama-3.3-70B

Dataset of confirmed self-correction episodes used for sequential activation analysis

Neighborhood — ranked by edge-count

Papers (1)

paper

Endogenous Resistance to Activation Steering in Language Models
introduces

Methods (1)

method

Sequential SAE Activation Analysis
uses
Token-level analysis of OTD and backtracking latent activations aligned at correction points across episodes