INDEX

Explanations

phrases related to negative consequences or criticism

oai_token-act-pair · gpt-3.5-turbo

terms associated with negative behaviors and outcomes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ractor

-0.87

 GOODMAN

-0.81

anooga

-0.75

enfranch

-0.73

anchester

-0.73

ector

-0.70

rounder

-0.69

astered

-0.68

psey

-0.68

atem

-0.67

POSITIVE LOGITS

 rash

0.98

cffff

0.95

smanship

0.93

 foolish

0.88

quin

0.81

 trusting

0.78

ly

0.78

ness

0.77

 folly

0.75

 arrogance

0.74

Activations Density 0.015%

phrases related to negative consequences or criticism

terms associated with negative behaviors and outcomes

No Comments

No Known Activations