INDEX

Explanations

government or politically related terms, specifically focused on committees, intelligence, services, and oversight

oai_token-act-pair · gpt-3.5-turbo

references to government committees and intelligence agencies

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

onyms

-0.67

 oneself

-0.66

GD

-0.66

 Barker

-0.66

idious

-0.64

 nets

-0.64

icular

-0.64

erd

-0.63

 Boko

-0.62

imated

-0.61

POSITIVE LOGITS

 Subcommittee

0.93

 agre

0.86

sembly

0.85

 Committee

0.81

 Appropriations

0.80

 subcommittee

0.78

 Resolution

0.78

 Appeals

0.78

 Caucus

0.77

 Council

0.76

Activations Density 0.079%

government or politically related terms, specifically focused on committees, intelligence, services, and oversight

references to government committees and intelligence agencies

No Comments

No Known Activations