INDEX

Explanations

mentions of speaking or having conversations with others

oai_token-act-pair · gpt-3.5-turbo

instances of the word "spoke" in various contexts

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

packs

-0.63

pack

-0.62

 Solitaire

-0.61

 Schedule

-0.60

 Deposit

-0.55

 Pirate

-0.55

 perennial

-0.55

Els

-0.55

orate

-0.54

leep

-0.54

POSITIVE LOGITS

 spoke

3.35

 speaks

1.95

 speak

1.93

 talked

1.93

 spoken

1.73

 speaking

1.58

 Speak

1.41

 sounded

1.33

 consulted

1.33

 interviewed

1.32

Activations Density 0.013%

mentions of speaking or having conversations with others

instances of the word "spoke" in various contexts

No Comments

No Known Activations