INDEX

Explanations

capitalized abbreviations followed by numerical values

oai_token-act-pair · gpt-3.5-turbo

patterns or metrics related to processes and their outcomes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

hement

-0.99

aterasu

-0.73

ataka

-0.70

 destro

-0.70

xual

-0.70

bda

-0.69

ueller

-0.69

ĸļ

-0.68

 honoured

-0.67

 agre

-0.66

POSITIVE LOGITS

ccording

0.89

ÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂł

0.87

References

0.78

ÂłÂłÂłÂłÂłÂłÂłÂł

0.73

>>>>>>>>

0.72

SOURCE

0.71

======

0.70

Marginal

0.70

Appearances

0.70

Dist

0.69

Activations Density 0.524%

capitalized abbreviations followed by numerical values

patterns or metrics related to processes and their outcomes

No Comments

No Known Activations