INDEX

Explanations

references to scoring in a sports context, particularly touchdowns

oai_token-act-pair · gpt-3.5-turbo

references to touchdowns in football

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

tan

-0.89

arest

-0.85

lly

-0.77

utical

-0.71

ename

-0.69

hemat

-0.69

emporary

-0.68

ħĭ

-0.68

States

-0.65

ruff

-0.65

POSITIVE LOGITS

 touchdown

1.08

 touchdowns

1.03

 receptions

0.85

 scored

0.83

 scoring

0.82

 rusher

0.79

 scorer

0.79

 clinch

0.74

 apiece

0.73

 reception

0.72

Activations Density 0.013%

references to scoring in a sports context, particularly touchdowns

references to touchdowns in football

No Comments

No Known Activations