INDEX

Explanations

negations with a focus on negative outcomes or potential problems

oai_token-act-pair · gpt-3.5-turbo

negative phrases indicating uncertainty or unresolved issues

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

lance

-0.73

alez

-0.69

ixel

-0.69

 ÂłÂł

-0.69

vant

-0.68

alid

-0.68

Shares

-0.67

ierrez

-0.64

riber

-0.64

plates

-0.62

POSITIVE LOGITS

 outright

1.09

 necessarily

1.02

 downright

0.96

 entirely

0.81

 altogether

0.78

ifies

0.74

 impossible

0.73

 quite

0.73

 eliminated

0.73

eworthy

0.72

Activations Density 0.035%

negations with a focus on negative outcomes or potential problems

negative phrases indicating uncertainty or unresolved issues

No Comments

No Known Activations