INDEX

Explanations

phrases related to confessions or admissions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "admit" or its variations

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

prot

-0.66

????????

-0.66

bench

-0.65

 grid

-0.62

map

-0.61

cycle

-0.61

 shape

-0.61

 ripple

-0.61

lance

-0.60

rend

-0.60

POSITIVE LOGITS

 admitted

3.59

 admits

2.30

 confessed

2.29

 admit

2.19

 acknowledged

2.02

 conceded

2.01

 admitting

1.98

 admission

1.95

 admissions

1.65

 concedes

1.53

Activations Density 0.008%

phrases related to confessions or admissions

instances of the word "admit" or its variations

No Comments

No Known Activations