INDEX

Explanations

phrases related to watching something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "watched."

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

olution

-0.66

 grant

-0.65

 clar

-0.64

 attribution

-0.61

 Soft

-0.61

olog

-0.61

 artic

-0.60

 compromise

-0.59

 grants

-0.59

istics

-0.57

POSITIVE LOGITS

 watched

3.47

 listened

2.04

 watches

1.94

 watch

1.92

 watching

1.85

 witnessed

1.69

 Watching

1.59

 monitored

1.58

watching

1.55

 streamed

1.55

Activations Density 0.017%

phrases related to watching something

instances of the word "watched."

No Comments

No Known Activations