INDEX

Explanations

phrases indicating recent viewing or watching activity

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Negative Logits

oing

-0.75

á½

-0.71

ogle

-0.70

Î³

-0.69

Reviewer

-0.69

Initialized

-0.68

ÏĢ

-0.66

alez

-0.66

ated

-0.65

Õ

-0.64

POSITIVE LOGITS

 WATCHED

1.25

Reply

0.65

trust

0.64

IELD

0.63

ifty

0.60

 Cabin

0.59

 âĢİ

0.59

 Continued

0.57

 Flake

0.56

Ago

0.55

Activations Density 0.093%

phrases indicating recent viewing or watching activity

No Known Activations