INDEX

Explanations

phrases related to emotions and opinions

oai_token-act-pair · gpt-3.5-turbo

negative contractions, particularly related to unwillingness or refusal

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 mixed

-0.62

 Windsor

-0.62

 SERV

-0.62

 soph

-0.60

 contrasted

-0.60

 Friends

-0.59

 fused

-0.59

 shack

-0.58

 blurred

-0.58

 partners

-0.57

POSITIVE LOGITS

1.09

nt

0.94

else

0.92

¹

0.90

ivably

0.87

ttle

0.87

erest

0.87

onna

0.85

swer

0.83

be

0.83

Activations Density 0.078%

phrases related to emotions and opinions

negative contractions, particularly related to unwillingness or refusal

No Comments

No Known Activations