INDEX

Explanations

text related to lists, rankings, and compilations

oai_token-act-pair · gpt-3.5-turbo

mentions of significant artworks or institutions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 compromise

-1.00

 stride

-1.00

 regulator

-0.98

 chancellor

-0.96

 slump

-0.95

 inhibitor

-0.94

rower

-0.94

 blaster

-0.93

 racer

-0.93

 flush

-0.92

POSITIVE LOGITS

 Characters

1.87

 Places

1.87

 Sites

1.86

 Lists

1.83

 Dates

1.83

 Languages

1.80

 Locations

1.78

 Cases

1.78

 Statements

1.77

 Programs

1.75

Activations Density 0.304%

text related to lists, rankings, and compilations

mentions of significant artworks or institutions

No Comments

No Known Activations