INDEX

Explanations

words related to research or confirmation of information

oai_token-act-pair · gpt-3.5-turbo

the concept of evidence or demonstrated effectiveness

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

doors

-0.72

 Dispatch

-0.66

Â§

-0.66

Boss

-0.66

reddits

-0.65

Tot

-0.64

door

-0.63

ignore

-0.62

atta

-0.61

Parts

-0.60

POSITIVE LOGITS

 proven

3.72

proven

2.53

 proved

1.92

 prove

1.51

 proving

1.49

 validated

1.35

 demonstrated

1.34

tested

1.29

 proves

1.24

 debunked

1.24

Activations Density 0.011%

words related to research or confirmation of information

the concept of evidence or demonstrated effectiveness

No Comments

No Known Activations