INDEX

Explanations

words associated with formal endorsements or validations

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

ivism

-0.79

iveness

-0.68

astical

-0.65

otom

-0.64

åĭ

-0.63

iop

-0.63

otten

-0.61

Balt

-0.60

ives

-0.60

uci

-0.60

POSITIVE LOGITS

gement

0.70

Reviewer

0.68

 Signed

0.68

Hash

0.65

 pathological

0.62

cedes

0.61

ename

0.61

odore

0.61

emption

0.60

 Flavoring

0.60

Activations Density 0.079%

words associated with formal endorsements or validations

No Comments

No Known Activations