INDEX

Explanations

phrases related to specific locations or entities, possibly related to politics, news, and potentially international matters

oai_token-act-pair · gpt-3.5-turbo

phrases related to academic institutions and affiliations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 hindsight

-0.73

 spoiler

-0.59

 inaction

-0.59

 improv

-0.56

 Extras

-0.50

 slang

-0.49

 Armageddon

-0.49

 scale

-0.48

 inconsistencies

-0.48

 Reloaded

-0.47

POSITIVE LOGITS

 pione

0.89

ù

0.75

Ě

0.75

ě

0.74

InstoreAndOnline

0.74

reportprint

0.74

ø

0.74

À

0.74

 RandomRedditor

0.74

Activations Density 1.492%

phrases related to specific locations or entities, possibly related to politics, news, and potentially international matters

phrases related to academic institutions and affiliations

No Comments

No Known Activations