INDEX

Explanations

important names or entities, likely related to a specific topic or person named ''anton''

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of specific names or identifiers, particularly "anton."

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

 bloated

-0.67

 crabs

-0.63

 shorts

-0.63

 Oscars

-0.62

opol

-0.62

 bundled

-0.62

Sri

-0.60

 Sponge

-0.59

 Buffy

-0.59

POSITIVE LOGITS

anton

4.11

oln

1.78

essen

1.08

raine

1.05

 surve

1.04

alon

0.93

 Garrison

0.88

 Myster

0.83

yden

0.82

ompl

0.80

Activations Density 0.026%

important names or entities, likely related to a specific topic or person named ''anton''

repeated mentions of specific names or identifiers, particularly "anton."

No Comments

No Known Activations