INDEX

Explanations

scientific abstracts or technical terms

oai_token-act-pair · gpt-3.5-turbo

references to academic abstracts or summaries

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 hate

-0.73

tip

-0.71

 sworn

-0.71

 unofficial

-0.70

 terror

-0.69

Tro

-0.64

ear

-0.63

rom

-0.63

TW

-0.62

 gearing

-0.62

POSITIVE LOGITS

Abstract

4.55

 Abstract

1.83

doi

1.61

stract

1.24

Synopsis

1.12

Summary

1.11

Auth

1.09

adobe

1.08

Researchers

1.02

GROUND

1.01

Activations Density 0.031%

scientific abstracts or technical terms

references to academic abstracts or summaries

No Comments

No Known Activations