INDEX

Explanations

conjunctions and phrases related to evaluating or comparing elements

oai_token-act-pair · gpt-3.5-turbo

phrases emphasizing conditionality and distinctions between permitted and prohibited actions

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

grand

-0.65

rocket

-0.65

 crim

-0.61

crim

-0.56

 Rover

-0.56

enne

-0.56

staking

-0.56

Detailed

-0.54

cision

-0.53

FORE

-0.53

POSITIVE LOGITS

how

1.15

 what

1.12

why

1.07

 whats

0.99

what

0.98

 whence

0.95

why

0.94

 where

0.92

WHY

0.91

 WHAT

0.88

Activations Density 0.079%

conjunctions and phrases related to evaluating or comparing elements

phrases emphasizing conditionality and distinctions between permitted and prohibited actions

No Comments

No Known Activations