INDEX

Explanations

instances of harmful behaviors or acts of aggression

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.10.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Negative Logits

ãĥ¼ãĥ³

-0.77

 Ended

-0.74

mother

-0.73

enders

-0.67

availability

-0.66

 emerges

-0.65

ends

-0.64

 arrived

-0.63

 arrives

-0.63

çļ

-0.63

POSITIVE LOGITS

 conversation

1.11

 discussions

1.04

 conversations

1.04

 activities

1.03

 vigorous

0.97

 productive

0.97

 dialogue

0.96

 hostilities

0.94

 risky

0.92

 discussion

0.92

Activations Density 0.373%

instances of harmful behaviors or acts of aggression

No Known Activations