INDEX

Explanations

the word "steel" at varying activations levels

oai_token-act-pair · gpt-3.5-turbo

references to steel and its various uses in industry and construction

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Kard

-0.74

uate

-0.70

itia

-0.70

olulu

-0.69

 Petr

-0.65

 Garr

-0.64

endment

-0.64

 Parties

-0.61

arios

-0.61

romeda

-0.61

POSITIVE LOGITS

 wool

1.14

 steel

1.00

worker

0.92

Series

0.90

 beams

0.87

Steel

0.85

anguage

0.85

steel

0.83

 oxide

0.82

workers

0.82

Activations Density 0.009%

the word "steel" at varying activations levels

references to steel and its various uses in industry and construction

No Comments

No Known Activations