INDEX

Explanations

sentences containing negations, particularly focusing on negations with high confidence

oai_token-act-pair · gpt-3.5-turbo

negations and phrases indicating something is not true or does not exist

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

arta

-0.87

=-=-=-=-=-=-=-=-

-0.77

 VIDEOS

-0.69

Dialogue

-0.68

 newsletters

-0.68

anon

-0.66

mare

-0.66

ologies

-0.66

NOW

-0.63

 certs

-0.63

POSITIVE LOGITS

 conceived

1.04

 unsuccessful

1.04

 originally

1.02

 instrumental

0.99

 initially

0.98

 born

0.93

 successful

0.91

 intended

0.87

 able

0.86

 supposed

0.83

Activations Density 0.283%

sentences containing negations, particularly focusing on negations with high confidence

negations and phrases indicating something is not true or does not exist

No Comments

No Known Activations