INDEX

Explanations

phrases related to news headlines and book titles

oai_token-act-pair · gpt-3.5-turbo

titles of media and references to their content

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

 periphery

-0.92

 linkage

-0.87

 suspension

-0.84

 prohibition

-0.82

 prohibitions

-0.81

 optics

-0.81

 caution

-0.80

 rendering

-0.80

 restraint

-0.80

 corridor

-0.79

POSITIVE LOGITS

 Changed

1.48

Own

1.43

 Wrong

1.42

 Worse

1.37

 Happ

1.36

 Alone

1.35

 Different

1.34

 Been

1.33

 Believe

1.33

 Killed

1.32

Activations Density 0.236%

phrases related to news headlines and book titles

titles of media and references to their content

No Comments

No Known Activations