INDEX

Explanations

expressions emphasizing intensity or emphasis

oai_token-act-pair · gpt-3.5-turbo

emphatic modifiers that indicate a strong degree of certainty or absolute statements

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

Not in Any Lists

No Comments

Negative Logits

yip

-0.91

rers

-0.80

anwhile

-0.79

olester

-0.76

ourses

-0.72

sburg

-0.72

*/(

-0.72

ulators

-0.72

EStream

-0.71

achu

-0.70

POSITIVE LOGITS

 understandable

1.04

 predictable

0.88

 justified

0.87

 unrelated

0.85

 unnecessary

0.82

 unacceptable

0.81

 antit

0.80

 different

0.78

 coinc

0.78

 legit

0.78

Activations Density 0.046%

expressions emphasizing intensity or emphasis

emphatic modifiers that indicate a strong degree of certainty or absolute statements

No Comments

No Known Activations