INDEX

Explanations

dates and times

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "at" followed by time or date references

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

aceutical

-0.79

employment

-0.66

userc

-0.65

llor

-0.65

irez

-0.65

HTTP

-0.65

gans

-0.64

vec

-0.64

film

-0.62

umatic

-0.62

POSITIVE LOGITS

 noon

1.15

 midnight

1.08

 halftime

1.05

 dusk

1.03

rium

1.02

 Wembley

0.98

 least

0.96

 midday

0.96

 sunset

0.89

 Disneyland

0.87

Activations Density 0.115%

dates and times

occurrences of the word "at" followed by time or date references

No Comments

No Known Activations