INDEX

Explanations

phrase highlighting a specific non-null token following an uncommon sequence 'ĊĊ'

oai_token-act-pair · gpt-3.5-turbo

specific numerical and categorical identifiers

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

•EggList

No Comments

Negative Logits

ģ«

-0.83

terday

-0.72

 predec

-0.70

ober

-0.70

oba

-0.70

omet

-0.68

ioned

-0.68

omical

-0.68

letal

-0.68

-0.66

POSITIVE LOGITS

1.12

sg

1.08

SR

1.06

Sak

1.02

SA

1.01

SS

0.98

 Samp

0.97

sa

0.96

Sv

0.96

0.95

Activations Density 0.323%

phrase highlighting a specific non-null token following an uncommon sequence 'ĊĊ'

specific numerical and categorical identifiers

No Comments

No Known Activations