INDEX

Explanations

words related to negative evaluations or criticism

oai_token-act-pair · gpt-3.5-turbo

words related to preference and taste

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

mers

-0.72

©¶æ

-0.68

ħĭ

-0.65

monds

-0.64

©¶æ¥µ

-0.63

assets

-0.62

 Haas

-0.62

 epid

-0.62

 crawl

-0.61

 ejac

-0.61

POSITIVE LOGITS

itism

1.20

ite

1.01

ited

0.96

naire

0.96

ITE

0.91

avour

0.89

eous

0.83

ably

0.83

 Favor

0.82

able

0.80

Activations Density 0.022%

words related to negative evaluations or criticism

words related to preference and taste

No Comments

No Known Activations