INDEX

Explanations

the word "clue" followed by either a single digit value or the word "clues" accompanied by a digit value expressing its strength of match

oai_token-act-pair · gpt-3.5-turbo

references to hints or pieces of evidence in various contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

violence

-0.68

 exceeding

-0.67

 conduct

-0.64

\/

-0.64

 Conduct

-0.61

 Motion

-0.61

 Expend

-0.60

Agg

-0.60

intensive

-0.59

 imposed

-0.58

POSITIVE LOGITS

 clue

3.73

 clues

2.82

 hint

1.73

 hints

1.54

cue

1.22

 cryptic

1.22

 cues

1.21

 decipher

1.17

 guesses

1.07

 answers

1.03

Activations Density 0.023%

the word "clue" followed by either a single digit value or the word "clues" accompanied by a digit value expressing its strength of match

references to hints or pieces of evidence in various contexts

No Comments

No Known Activations