INDEX

Explanations

terms related to prolonging or extending various concepts or conditions

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

leck

-0.72

illard

-0.66

ches

-0.66

ids

-0.63

te

-0.63

«

-0.62

imming

-0.62

CENT

-0.61

footed

-0.61

tics

-0.60

POSITIVE LOGITS

 warranties

0.86

 adolesc

0.82

 invitations

0.75

 beyond

0.73

 condolences

0.72

 cords

0.72

ISS

0.70

 lifespan

0.70

 Reach

0.69

ibility

0.68

Activations Density 0.055%

terms related to prolonging or extending various concepts or conditions

No Comments

No Known Activations