INDEX

Explanations

references to the word "grass" at varying activations

oai_token-act-pair · gpt-3.5-turbo

references to grass

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

icts

-0.73

osexual

-0.69

umbn

-0.69

 Somers

-0.66

 Heller

-0.65

andem

-0.64

 Siem

-0.64

ISM

-0.64

 critically

-0.63

 prefrontal

-0.62

POSITIVE LOGITS

roots

1.19

lands

1.17

ho

1.14

wood

1.02

flower

1.00

banks

0.95

nut

0.91

woods

0.90

 airst

0.90

land

0.89

Activations Density 0.016%

references to the word "grass" at varying activations

references to grass

No Comments

No Known Activations