INDEX

Explanations

references to news organizations and possibly crime-related terms

oai_token-act-pair · gpt-3.5-turbo

references to specific news outlets and events involving individuals or testimonies in reports

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

itance

-0.84

utor

-0.82

 gener

-0.75

ometric

-0.74

utory

-0.74

table

-0.71

depend

-0.70

arial

-0.69

sed

-0.68

idas

-0.68

POSITIVE LOGITS

CBS

1.02

radio

1.01

 radio

1.00

 Radio

0.98

Radio

0.95

ILLE

0.90

TV

0.90

Channel

0.87

 Philly

0.87

 Television

0.86

Activations Density 0.119%

references to news organizations and possibly crime-related terms

references to specific news outlets and events involving individuals or testimonies in reports

No Comments

No Known Activations