INDEX

Explanations

adjectives related to size or length

oai_token-act-pair · gpt-3.5-turbo

references to short duration or brevity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

velt

-0.71

uth

-0.70

uin

-0.66

enger

-0.62

udic

-0.61

grown

-0.61

igators

-0.61

boxes

-0.60

adata

-0.59

 Bros

-0.59

POSITIVE LOGITS

 short

3.57

short

2.43

Short

2.26

 Short

2.26

 shorter

2.06

 brief

1.88

 shortest

1.71

 shorts

1.68

 shortened

1.54

 shorten

1.51

Activations Density 0.028%

adjectives related to size or length

references to short duration or brevity

No Comments

No Known Activations