INDEX

Explanations

names or terms associated with legal cases or political figures

oai_token-act-pair · gpt-3.5-turbo

the term "bar" 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

NF

-0.68

 seeded

-0.66

ispers

-0.65

 Participant

-0.64

 Ghostbusters

-0.63

 elig

-0.63

 transcripts

-0.62

ITNESS

-0.61

LECT

-0.61

 genuine

-0.61

POSITIVE LOGITS

bar

4.41

bars

3.15

Bar

2.10

bar

1.90

Bar

1.66

 Bars

1.52

 bars

1.48

bart

1.43

beer

1.38

BAR

1.33

Activations Density 0.009%

names or terms associated with legal cases or political figures

the term "bar" in various contexts

No Comments

No Known Activations