INDEX

Explanations

highly cited information or references in a text

oai_token-act-pair · gpt-3.5-turbo

phrases related to academic citations

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

wards

-0.83

 Stretch

-0.71

achine

-0.71

locks

-0.70

oops

-0.70

 goodbye

-0.66

ritch

-0.65

 halves

-0.65

OOOO

-0.65

 goodness

-0.65

POSITIVE LOGITS

 cited

3.67

 quoted

2.28

 cite

2.12

 cites

2.08

 citing

1.91

 referenced

1.86

 citations

1.74

 citation

1.63

 credited

1.52

 touted

1.51

Activations Density 0.014%

highly cited information or references in a text

phrases related to academic citations

No Comments

No Known Activations