INDEX

Explanations

terms related to time sequences or ordering

oai_token-act-pair · gpt-3.5-turbo

the word "Next" and its variations, indicating a focus on sequential or upcoming events

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

 divers

-0.75

 sometimes

-0.71

 recorded

-0.68

fr

-0.66

 plain

-0.65

 stem

-0.64

 warrant

-0.62

 hurt

-0.61

 medically

-0.61

 appearing

-0.61

POSITIVE LOGITS

Next

3.65

next

1.88

Secondly

1.72

 Next

1.72

Finally

1.56

Then

1.52

Lastly

1.45

Tomorrow

1.41

Later

1.41

Moving

1.41

Activations Density 0.020%

terms related to time sequences or ordering

the word "Next" and its variations, indicating a focus on sequential or upcoming events

No Comments

No Known Activations