INDEX

Explanations

technical abbreviations or acronyms ending in 'AC'

oai_token-act-pair · gpt-3.5-turbo

references to organizational or agency acronyms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Fenrir

-0.72

hood

-0.72

 Colt

-0.72

backer

-0.66

 Veter

-0.66

Mub

-0.63

Pax

-0.60

 vomit

-0.60

 Dolphin

-0.59

ãĥĥãĥĪ

-0.59

POSITIVE LOGITS

CEPT

1.05

HE

1.04

olor

1.00

rylic

0.98

osta

0.94

ritical

0.94

ORN

0.92

hetto

0.89

oustic

0.89

RAFT

0.88

Activations Density 0.006%

technical abbreviations or acronyms ending in 'AC'

references to organizational or agency acronyms

No Comments

No Known Activations