INDEX

Explanations

phrases containing coordinated entities, such as organizations or government agencies

oai_token-act-pair · gpt-3.5-turbo

instances of the word "and" to identify conjunctions linking multiple subjects or ideas

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

éŃĶ

-0.73

objects

-0.73

ãĥ¼ãĥ

-0.72

animate

-0.70

ivari

-0.69

mol

-0.68

spir

-0.67

bo

-0.67

å¦

-0.67

æĥ

-0.65

POSITIVE LOGITS

 vice

0.88

 affiliated

0.87

 colleagues

0.85

 allied

0.84

 staffed

0.83

 headquartered

0.81

 assorted

0.79

 Associates

0.78

 academia

0.75

 associates

0.75

Activations Density 0.346%

phrases containing coordinated entities, such as organizations or government agencies

instances of the word "and" to identify conjunctions linking multiple subjects or ideas

No Comments

No Known Activations