INDEX

Explanations

phrases associated with giving instructions or commands

oai_token-act-pair · gpt-3.5-turbo

informal conversational transitions or prompts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

âĵĺ

-0.78

nai

-0.74

cit

-0.73

iddler

-0.72

mination

-0.70

phant

-0.69

mone

-0.69

ache

-0.67

³

-0.67

ROR

-0.66

POSITIVE LOGITS

 Lets

0.97

let

0.95

 alright

0.95

 congratulations

0.95

bye

0.94

 okay

0.93

 maybe

0.91

LET

0.91

 lets

0.91

 Okay

0.90

Activations Density 0.123%

phrases associated with giving instructions or commands

informal conversational transitions or prompts

No Comments

No Known Activations