INDEX

Explanations

quantitative ratios such as "seven out of ten."

oai_token-act-pair · gpt-3.5-turbo

phrases indicating proportions or statistics expressed in fractional terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

tyr

-0.86

 arsen

-0.63

Yen

-0.57

 background

-0.54

 Elements

-0.54

Der

-0.51

BLE

-0.51

ITED

-0.51

 Bulgar

-0.50

 Nadu

-0.50

POSITIVE LOGITS

lier

1.05

ta

0.98

liest

0.94

number

0.92

fitted

0.90

ranked

0.88

lander

0.88

of

0.87

lying

0.87

paced

0.87

Activations Density 0.023%

quantitative ratios such as "seven out of ten."

phrases indicating proportions or statistics expressed in fractional terms

No Comments

No Known Activations