INDEX

Explanations

phrases indicating recommended actions or opinions on actions to be taken

oai_token-act-pair · gpt-3.5-turbo

expressions of recommendation or obligation

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

 cryptic

-0.65

 cycles

-0.65

Pse

-0.64

 slips

-0.64

 unlucky

-0.61

Cir

-0.61

Sci

-0.60

 Bonds

-0.58

WI

-0.57

ãĤ¼

-0.57

POSITIVE LOGITS

 reconsider

0.96

©¶æ

0.92

 rethink

0.90

 apologise

0.88

 ashamed

0.84

eryl

0.81

 apologize

0.79

 clarify

0.79

ople

0.79

rador

0.78

Activations Density 0.219%

phrases indicating recommended actions or opinions on actions to be taken

expressions of recommendation or obligation

No Comments

No Known Activations