INDEX

Explanations

percentage statistics and survey respondents

oai_token-act-pair · gpt-3.5-turbo

mentions of survey participants or respondents

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 sorcery

-0.71

inside

-0.66

 Fighting

-0.61

 fuse

-0.61

 sneak

-0.60

unity

-0.59

 plane

-0.59

Magic

-0.59

 Plane

-0.57

alt

-0.57

POSITIVE LOGITS

 respondents

3.61

 respondent

2.45

 Respond

1.70

 participants

1.64

 applicants

1.62

 surveyed

1.57

 commenters

1.55

 attendees

1.44

 polled

1.43

 entrants

1.39

Activations Density 0.019%

percentage statistics and survey respondents

mentions of survey participants or respondents

No Comments

No Known Activations