INDEX

Explanations

phrases related to questions or instructions with multiple choices

oai_token-act-pair · gpt-3.5-turbo

phrases beginning with a forward slash (/), likely indicating topics or categories

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Fletcher

-0.67

Syl

-0.66

 Carlton

-0.64

Dug

-0.64

 Pigs

-0.63

 warr

-0.63

 Giov

-0.63

 avail

-0.63

 creep

-0.62

agascar

-0.61

POSITIVE LOGITS

etc

1.34

whatever

1.11

pmwiki

1.04

pol

0.98

0.97

coll

0.96

Getty

0.96

oct

0.95

AFP

0.95

workshop

0.94

Activations Density 0.094%

phrases related to questions or instructions with multiple choices

phrases beginning with a forward slash (/), likely indicating topics or categories

No Comments

No Known Activations