INDEX

Explanations

capitalized words that appear to be related to locations or proper names

oai_token-act-pair · gpt-3.5-turbo

mentions of "Ta" or related terms that may indicate a geographical or cultural reference

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

-0.71

CTR

-0.70

CRC

-0.70

olitan

-0.65

Ö¼

-0.63

 mapped

-0.63

 Geographic

-0.60

PNG

-0.60

 Continental

-0.60

 Governments

-0.59

POSITIVE LOGITS

Ta

1.08

plin

1.03

iba

0.91

ween

0.89

Ta

0.86

fts

0.81

oshi

0.80

ft

0.79

raq

0.78

vern

0.77

Activations Density 0.006%

capitalized words that appear to be related to locations or proper names

mentions of "Ta" or related terms that may indicate a geographical or cultural reference

No Comments

No Known Activations