INDEX

Explanations

references to time durations in weeks

oai_token-act-pair · gpt-3.5-turbo

references to time intervals, specifically weeks

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

 power

-0.68

 transmission

-0.63

 protected

-0.63

 corruption

-0.62

 descendants

-0.61

 inherited

-0.61

 bios

-0.60

 fame

-0.60

 swearing

-0.60

 shields

-0.59

POSITIVE LOGITS

week

3.87

month

2.30

Week

2.27

day

1.98

 Week

1.83

 WEEK

1.81

 week

1.78

weekly

1.58

morning

1.56

Month

1.55

Activations Density 0.012%

references to time durations in weeks

references to time intervals, specifically weeks

No Comments

No Known Activations