INDEX

Explanations

numerical information or lists related to various subjects

oai_token-act-pair · gpt-3.5-turbo

quantitative information related to rankings or lists

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 disagreed

-0.67

 methodological

-0.66

techn

-0.64

 procedural

-0.63

sych

-0.62

 Controls

-0.62

pmwiki

-0.61

erness

-0.61

 inconsist

-0.60

 inacc

-0.60

POSITIVE LOGITS

 coffers

1.15

 lucrative

1.13

 millionaires

1.12

 millionaire

1.11

 riches

1.09

 prizes

1.08

 auctions

1.08

 profits

1.07

 payout

1.06

 cash

1.03

Activations Density 0.865%

numerical information or lists related to various subjects

quantitative information related to rankings or lists

No Comments

No Known Activations