INDEX

Explanations

words related to completing a task or reaching a conclusion

oai_token-act-pair · gpt-3.5-turbo

instances of the word "finish" or its variations in various contexts

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

rays

-0.72

atri

-0.68

 fost

-0.65

CTV

-0.65

quest

-0.64

cult

-0.64

agents

-0.63

Mand

-0.63

Wars

-0.63

embed

-0.63

POSITIVE LOGITS

 finish

3.90

 Finish

2.71

 finishes

2.66

 finishing

2.36

Finish

2.32

 finished

2.01

 completion

1.41

finished

1.37

 Finished

1.36

fin

1.15

Activations Density 0.013%

words related to completing a task or reaching a conclusion

instances of the word "finish" or its variations in various contexts

No Comments

No Known Activations