INDEX

Explanations

various mix of letters and symbols with sporadic activations

oai_token-act-pair · gpt-3.5-turbo

abbreviations or shorthand representations, possibly related to identifiers or codes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 today

-0.60

redibly

-0.54

 ÂŃ

-0.54

 shockingly

-0.51

 reminder

-0.51

 batter

-0.49

 homicide

-0.49

INGTON

-0.48

 opposite

-0.48

 rented

-0.48

POSITIVE LOGITS

ZX

1.11

==

1.11

=/

1.06

1.00

Ul

0.99

XM

0.98

Ry

0.97

dq

0.96

/+

0.95

Iv

0.95

Activations Density 0.027%

various mix of letters and symbols with sporadic activations

abbreviations or shorthand representations, possibly related to identifiers or codes

No Comments

No Known Activations