INDEX

Explanations

highlights or featured content

oai_token-act-pair · gpt-3.5-turbo

mentions of highlights or notable features from various content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 heal

-0.79

 healing

-0.70

act

-0.70

hel

-0.69

 unin

-0.67

 calculating

-0.66

 webs

-0.66

 tract

-0.64

unt

-0.63

 wound

-0.62

POSITIVE LOGITS

 Highlights

4.24

 highlights

1.88

 highlight

1.19

 Preview

1.17

 Features

1.08

 Recap

1.08

 Coverage

1.01

 Details

0.99

 Overview

0.98

 Summary

0.97

Activations Density 0.005%

highlights or featured content

mentions of highlights or notable features from various content

No Comments

No Known Activations