INDEX

Explanations

phrases related to rankings, positions, and competitions

oai_token-act-pair · gpt-3.5-turbo

sequences that indicate rankings or placements in a competitive context

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

``

-0.84

assian

-0.79

-0.75

ards

-0.73

ega

-0.72

 Rican

-0.71

vision

-0.71

stra

-0.65

opian

-0.65

 Replay

-0.64

POSITIVE LOGITS

————

4.34

————————

3.64

————————————————

3.31

——

2.65

âĢ¦âĢ¦âĢ¦âĢ¦

1.83

âĢ¦âĢ¦âĢ¦âĢ¦âĢ¦âĢ¦âĢ¦âĢ¦

1.81

—-

1.50

****************

1.38

âĢ¢âĢ¢âĢ¢âĢ¢

1.34

********************************

1.29

Activations Density 0.048%

phrases related to rankings, positions, and competitions

sequences that indicate rankings or placements in a competitive context

No Comments

No Known Activations