INDEX

Explanations

phrases related to Q&A sessions

oai_token-act-pair · gpt-3.5-turbo

instances of question and answer (Q&A) formats and the character "&"

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

ufact

-0.81

heed

-0.74

raint

-0.74

gins

-0.68

fell

-0.66

 fore

-0.61

bourg

-0.60

 orche

-0.59

wolf

-0.58

etheless

-0.58

POSITIVE LOGITS

0.95

Answer

0.87

oS

0.87

FAQ

0.86

Args

0.82

0.81

reply

0.81

Reply

0.80

Response

0.79

AMA

0.79

Activations Density 0.032%

phrases related to Q&A sessions

instances of question and answer (Q&A) formats and the character "&"

No Comments

No Known Activations