INDEX

Explanations

negative statements or skepticism

oai_token-act-pair · gpt-3.5-turbo

negative phrases, particularly with the word "not."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

accompan

-0.61

 Opportun

-0.56

 Dynamics

-0.55

 Exploration

-0.54

 nonetheless

-0.54

 Enhanced

-0.54

 whichever

-0.53

 alternatively

-0.53

 avenues

-0.53

arthy

-0.53

POSITIVE LOGITS

hin

0.92

 kidding

0.91

 disappoint

0.86

 anymore

0.81

 underestimate

0.80

 condone

0.80

 discriminate

0.76

 tolerate

0.74

 hesitate

0.73

 bothered

0.72

Activations Density 0.238%

negative statements or skepticism

negative phrases, particularly with the word "not."

No Comments

No Known Activations