INDEX

Explanations

numeric values structured in a specific pattern

oai_token-act-pair · gpt-3.5-turbo

numerical values, particularly those related to object counts or identifiers

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

 adaptation

-0.68

 grace

-0.67

 flexibility

-0.67

 maker

-0.66

 favoured

-0.62

 beauty

-0.61

gre

-0.61

 snapping

-0.60

 dominance

-0.60

 Ambrose

-0.58

POSITIVE LOGITS

3.94

2.99

2.92

2.75

2.43

00000000

2.40

2.11

1.99

1.95

Activations Density 0.014%

numeric values structured in a specific pattern

numerical values, particularly those related to object counts or identifiers

No Comments

No Known Activations