INDEX

Explanations

mentions of physical actions related to breaking or fracturing

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of the term "breaks."

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

omorphic

-0.73

Gur

-0.66

named

-0.66

jud

-0.65

nom

-0.65

aed

-0.64

 Poster

-0.64

erie

-0.64

Chan

-0.62

CENT

-0.62

POSITIVE LOGITS

 breaks

3.74

 break

2.34

breaks

2.32

 Break

1.70

 broke

1.69

break

1.68

Break

1.66

 breaking

1.47

 pauses

1.45

 bends

1.44

Activations Density 0.010%

mentions of physical actions related to breaking or fracturing

repeated mentions of the term "breaks."

No Comments

No Known Activations