INDEX

Explanations

Keywords related to programming and technical instructions

oai_token-act-pair · gpt-3.5-turbo

indicative phrases related to success or achievement

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

•Suppression Features Layer 8 GPT2 Small

No Comments

Negative Logits

-0.61

ē

-0.61

Ď

-0.61

ė

-0.61

ā

-0.61

ě

-0.61

 TheNitrome

-0.61

û

-0.61

 RandomRedditor

-0.61

rawdownload

-0.61

POSITIVE LOGITS

 blur

0.35

LLC

0.34

 supra

0.33

(~

0.33

 spat

0.33

 Byte

0.32

 synerg

0.32

(@

0.32

 Editor

0.32

 respectively

0.31

Activations Density 2.525%

Keywords related to programming and technical instructions

indicative phrases related to success or achievement

No Comments

No Known Activations