INDEX

Explanations

references to objects or items described as pieces or parts

oai_token-act-pair · gpt-3.5-turbo

terms related to artistic or conceptual works, often referred to as "pieces."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ELL

-0.73

 common

-0.65

 footh

-0.63

 Unified

-0.63

 developing

-0.61

 govern

-0.61

 nascent

-0.60

 popular

-0.59

 Kindle

-0.57

 pronounced

-0.57

POSITIVE LOGITS

meal

1.29

piece

1.23

pieces

1.14

OGR

0.94

ngth

0.90

ificate

0.89

ridges

0.89

glass

0.84

adoes

0.84

fare

0.83

Activations Density 0.009%

references to objects or items described as pieces or parts

terms related to artistic or conceptual works, often referred to as "pieces."

No Comments

No Known Activations