INDEX

Explanations

numerical values such as percentages or codes within a text

oai_token-act-pair · gpt-3.5-turbo

numerical values and statistical terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 resistor

-0.57

 late

-0.57

Wer

-0.55

 another

-0.55

 bribe

-0.54

 sometime

-0.53

 bullshit

-0.53

 wait

-0.53

 exceptions

-0.52

POSITIVE LOGITS

ãĤ«

0.80

Ùħ

0.77

¥ŀ

0.77

umbn

0.76

 teasp

0.75

ğ

0.72

InstoreAndOnline

0.72

č

0.72

 externalToEVA

0.72

ě

0.72

Activations Density 0.892%

numerical values such as percentages or codes within a text

numerical values and statistical terms

No Comments

No Known Activations