INDEX

Explanations

proper nouns followed by action verbs indicating success or achievement

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "won" associated with awards or achievements

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

zik

-0.63

 dstg

-0.63

Ħ¢

-0.62

LINE

-0.60

Newsletter

-0.59

grow

-0.58

 coales

-0.58

Alternatively

-0.57

Normal

-0.57

existing

-0.57

POSITIVE LOGITS

 acclaim

1.06

't

1.02

 accol

0.94

 medals

0.89

 awards

0.89

 rave

0.87

 reelection

0.86

 praise

0.84

 renown

0.84

 Rookie

0.83

Activations Density 0.081%

proper nouns followed by action verbs indicating success or achievement

occurrences of the word "won" associated with awards or achievements

No Comments

No Known Activations