INDEX

Explanations

small pieces of interesting information or details

oai_token-act-pair · gpt-3.5-turbo

mentions of "tidbits" or small pieces of information

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

pour

-0.75

terday

-0.74

irez

-0.70

AIN

-0.68

ONT

-0.63

hovah

-0.63

sein

-0.61

posure

-0.61

HAEL

-0.60

enson

-0.59

POSITIVE LOGITS

bits

1.04

ches

0.79

ewater

0.77

================================================================

0.76

istically

0.75

ching

0.72

ily

0.71

quo

0.70

âĶĢ

0.70

ying

0.70

Activations Density 0.111%

small pieces of interesting information or details

mentions of "tidbits" or small pieces of information

No Comments

No Known Activations