INDEX

Explanations

proper nouns and names related to individuals

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of specific names, particularly "Willis," "Thames," and "Jonas."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ebook

-0.96

ardless

-0.77

fecture

-0.74

utable

-0.74

bies

-0.74

uling

-0.74

urrent

-0.74

etus

-0.69

aby

-0.69

vette

-0.68

POSITIVE LOGITS

ophical

0.82

éĹĺ

0.74

Ô

0.74

\\\\\\\\

0.74

peed

0.72

sein

0.71

 Cause

0.70

 GOODMAN

0.69

paces

0.67

kinson

0.67

Activations Density 0.027%

proper nouns and names related to individuals

repeated mentions of specific names, particularly "Willis," "Thames," and "Jonas."

No Comments

No Known Activations