INDEX

Explanations

references to numerical values associated with rankings and statistics

oai_token-act-pair · gpt-3.5-turbo

forms of the verb "to be" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 externalToEVAOnly

-0.78

DragonMagazine

-0.77

oice

-0.66

aceae

-0.63

eral

-0.61

work

-0.59

obi

-0.59

imental

-0.59

ieval

-0.58

 misunderstand

-0.58

POSITIVE LOGITS

 averaging

1.05

 fourth

1.01

 trailing

0.96

 seventh

0.96

 eighth

0.94

 tied

0.94

 fifth

0.94

 sixth

0.94

 ninth

0.93

 beating

0.89

Activations Density 0.179%

references to numerical values associated with rankings and statistics

forms of the verb "to be" in various contexts

No Comments

No Known Activations