INDEX

Explanations

references to specific time periods, particularly weeks

oai_token-act-pair · gpt-3.5-turbo

references to "week" with varying contexts, particularly those that indicate timeframes

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hess

-0.76

rices

-0.65

its

-0.62

found

-0.62

rics

-0.62

Gra

-0.61

Edge

-0.60

endants

-0.60

 Ariel

-0.60

ius

-0.59

POSITIVE LOGITS

 week

3.30

 month

2.54

week

2.43

 Week

2.30

 fortnight

2.28

 weekend

2.24

 WEEK

2.19

Week

2.09

 weeks

2.08

 Month

1.98

Activations Density 0.041%

references to specific time periods, particularly weeks

references to "week" with varying contexts, particularly those that indicate timeframes

No Comments

No Known Activations