INDEX

Explanations

exclamations and affirmations

oai_token-act-pair · gpt-3.5-turbo

expressions of disagreement or negative responses

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

teenth

-1.05

velop

-0.77

icipated

-0.76

 ILCS

-0.74

irth

-0.73

States

-0.73

emouth

-0.73

ugal

-0.73

ertility

-0.73

 fullest

-0.71

POSITIVE LOGITS

 Nope

1.11

 blah

1.00

 damned

0.86

Yep

0.85

 darn

0.81

 Sorry

0.79

bye

0.77

 REALLY

0.77

 sorry

0.77

 dunno

0.76

Activations Density 0.031%

exclamations and affirmations

expressions of disagreement or negative responses

No Comments

No Known Activations