INDEX

Explanations

phrases related to past events or timeframes

oai_token-act-pair · gpt-3.5-turbo

references to time, specifically events mentioned with the word "last" followed by a numeral

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

cil

-0.80

holes

-0.75

kaya

-0.66

aband

-0.66

aunders

-0.65

gans

-0.62

 Franch

-0.61

ollow

-0.60

ular

-0.59

ikan

-0.59

POSITIVE LOGITS

 month

0.98

 week

0.97

 ditch

0.93

 year

0.90

 gasp

0.86

 night

0.86

 rites

0.80

 summer

0.78

 Month

0.78

 Gleaming

0.77

Activations Density 0.078%

phrases related to past events or timeframes

references to time, specifically events mentioned with the word "last" followed by a numeral

No Comments

No Known Activations