INDEX

Explanations

proper nouns, particularly names of individuals or titles

oai_token-act-pair · gpt-3.5-turbo

proper nouns and significant numerical values

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

Not in Any Lists

No Comments

Negative Logits

 Brawl

-0.83

 UNHCR

-0.80

 Quinn

-0.79

TAG

-0.77

 Schwarz

-0.75

Zub

-0.74

Wes

-0.74

Sync

-0.74

 Shack

-0.72

 Syndicate

-0.71

POSITIVE LOGITS

 Prize

0.80

gra

0.78

iffe

0.76

ARDS

0.76

[+

0.73

cause

0.71

 Alonso

0.71

 mustard

0.70

ifer

0.69

illa

0.69

Activations Density 0.519%

proper nouns, particularly names of individuals or titles

proper nouns and significant numerical values

No Comments

No Known Activations