INDEX

Explanations

the word "Simpson" with varying activations

oai_token-act-pair · gpt-3.5-turbo

the presence of the name "Simpson" and related terms associated with it in various contexts

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

 cess

-0.76

 Terra

-0.72

 Leilan

-0.70

 Guards

-0.68

perty

-0.68

 moth

-0.66

rescent

-0.66

 creepy

-0.64

 monarch

-0.63

 oily

-0.63

POSITIVE LOGITS

 Simpson

3.57

Sim

1.52

sim

1.19

 Eaton

1.07

 Simone

1.02

 Sterling

1.02

 Lennon

1.00

 Sinclair

0.98

 Kardashian

0.95

 Cutler

0.94

Activations Density 0.016%

the word "Simpson" with varying activations

the presence of the name "Simpson" and related terms associated with it in various contexts

No Comments

No Known Activations