INDEX

Explanations

technical or scientific terms and concepts

oai_token-act-pair · gpt-3.5-turbo

assertions and evaluations about possibilities and moral judgments

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

boarding

-0.81

enaries

-0.74

ventures

-0.74

 Tackle

-0.73

luaj

-0.72

licks

-0.70

rients

-0.68

scenes

-0.68

 Emails

-0.66

anamo

-0.65

POSITIVE LOGITS

 conceivable

1.34

 logically

1.26

 meaningless

1.22

 improbable

1.22

omorphic

1.19

 impossible

1.19

 spurious

1.18

 plausible

1.17

 irrelevant

1.15

 triv

1.13

Activations Density 0.235%

technical or scientific terms and concepts

assertions and evaluations about possibilities and moral judgments

No Comments

No Known Activations