INDEX

Explanations

questions and uncertainties related to societal and personal beliefs

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

Not in Any Lists

No Comments

Negative Logits

orage

-0.61

 therefore

-0.60

estamp

-0.59

ovember

-0.59

 Sabha

-0.59

ords

-0.55

ovation

-0.55

ãĥ£

-0.55

 importance

-0.54

eting

-0.53

POSITIVE LOGITS

 alternatively

1.11

abouts

1.05

phans

1.03

thodox

0.89

Else

0.81

 altern

0.79

depending

0.78

gans

0.78

å§«

0.74

 worse

0.73

Activations Density 0.527%

questions and uncertainties related to societal and personal beliefs

No Comments

No Known Activations