INDEX

Explanations

phrases indicating a statement or point being made

oai_token-act-pair · gpt-3.5-turbo

phrases emphasizing personal opinions or requests

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

 achievable

-0.51

OD

-0.51

 majorities

-0.51

 nowhere

-0.50

 okay

-0.49

OT

-0.49

OK

-0.48

 progress

-0.48

 playable

-0.48

 disadvant

-0.48

POSITIVE LOGITS

REDACTED

0.63

uel

0.62

stal

0.61

kel

0.60

ickle

0.59

culus

0.59

elo

0.58

lde

0.58

fell

0.58

FORE

0.57

Activations Density 0.432%

phrases indicating a statement or point being made

phrases emphasizing personal opinions or requests

No Comments

No Known Activations