INDEX

Explanations

phrases or words related to specific cultural or political references

oai_token-act-pair · gpt-3.5-turbo

references to "gags" or related comedic elements in the text

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

 mutual

-0.73

ccording

-0.71

 reconstruction

-0.63

 neurolog

-0.62

 genetic

-0.61

 simultaneous

-0.61

Nex

-0.60

 bilateral

-0.60

 mitochondrial

-0.58

 medicine

-0.58

POSITIVE LOGITS

ags

1.31

gers

0.97

gery

0.91

zag

0.90

agging

0.89

ansas

0.88

glers

0.88

gered

0.86

aws

0.85

weed

0.84

Activations Density 0.004%

phrases or words related to specific cultural or political references

references to "gags" or related comedic elements in the text

No Comments

No Known Activations