INDEX

Explanations

text that appear to be computer code or technical in nature

oai_token-act-pair · gpt-3.5-turbo

specific capitalized acronyms or alphanumeric sequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 derog

-0.65

 mechanically

-0.63

 sabotage

-0.63

 instruments

-0.62

 vaguely

-0.61

 vacuum

-0.60

 indications

-0.60

 briefly

-0.58

 refres

-0.58

 cardboard

-0.57

POSITIVE LOGITS

ZX

1.14

XM

1.05

Iv

1.05

CN

1.01

Ct

0.99

Fu

0.99

Rh

0.98

Ry

0.98

=/

0.98

Ni

0.97

Activations Density 0.039%

text that appear to be computer code or technical in nature

specific capitalized acronyms or alphanumeric sequences

No Comments

No Known Activations