INDEX

Explanations

terms related to holding or putting something on hold

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Hold" and variations of it

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

erson

-0.71

 idol

-0.66

 degrading

-0.64

 transforming

-0.63

 decaying

-0.63

 wasteful

-0.63

 tart

-0.62

 phosphorus

-0.62

 union

-0.62

 decomp

-0.62

POSITIVE LOGITS

 Hold

3.87

Hold

2.39

hold

1.73

 Held

1.61

 Holding

1.47

 Holder

1.44

 hold

1.31

holding

1.18

 Knock

1.18

 Reach

1.17

Activations Density 0.018%

terms related to holding or putting something on hold

instances of the word "Hold" and variations of it

No Comments

No Known Activations