INDEX

Explanations

government-related terms, specifically referring to departments or agencies

oai_token-act-pair · gpt-3.5-turbo

references to government departments and organizations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 bidder

-0.79

 bandwagon

-0.73

 smoker

-0.72

 hawk

-0.71

 slate

-0.71

 heap

-0.71

 gamer

-0.67

 believer

-0.67

 bubble

-0.67

 manufacturer

-0.67

POSITIVE LOGITS

 Affairs

1.27

 Relations

0.99

 Vehicles

0.97

 Regulation

0.88

 Responsibility

0.86

 Activities

0.85

 Services

0.85

 Controls

0.80

 Instruction

0.79

Services

0.79

Activations Density 0.133%

government-related terms, specifically referring to departments or agencies

references to government departments and organizations

No Comments

No Known Activations