INDEX

Explanations

reviewers or commenters

oai_token-act-pair · gpt-3.5-turbo

references to individuals involved in reviewing or evaluating content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Centers

-0.72

 Medic

-0.66

yles

-0.65

 Pesh

-0.64

rams

-0.63

spring

-0.63

Medic

-0.62

lesiastical

-0.61

 Lent

-0.61

 Pump

-0.60

POSITIVE LOGITS

Reviewer

1.11

 reviewers

1.10

 reviewer

0.94

Reviewed

0.90

Redditor

0.74

ikawa

0.73

iffs

0.72

 reviews

0.70

untled

0.69

 eager

0.69

Activations Density 0.014%

reviewers or commenters

references to individuals involved in reviewing or evaluating content

No Comments

No Known Activations