INDEX

Explanations

contractions with "n't" and variations of the word "not"

oai_token-act-pair · gpt-3.5-turbo

negations or phrases indicating disagreement or denial

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 guessing

-0.84

 glad

-0.76

 spirited

-0.75

 unable

-0.67

 lucky

-0.66

estern

-0.66

 unwilling

-0.65

holding

-0.65

 nervous

-0.65

 restless

-0.65

POSITIVE LOGITS

 necessarily

1.24

 preclude

1.18

 correlate

1.08

 detract

1.08

 constitute

1.06

 suffice

1.06

 affect

1.05

 occur

1.03

 coincide

1.01

 imply

0.98

Activations Density 0.167%

contractions with "n't" and variations of the word "not"

negations or phrases indicating disagreement or denial

No Comments

No Known Activations