INDEX

Explanations

words related to recognition or honor

oai_token-act-pair · gpt-3.5-turbo

references to the name "John" and variations of its spelling

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Tube

-0.76

 Tube

-0.75

 needles

-0.75

 panicked

-0.74

 blob

-0.71

 Panic

-0.70

 chunks

-0.70

 tube

-0.70

 dense

-0.69

POSITIVE LOGITS

hon

3.60

 Honor

3.11

Hon

2.87

 Honour

2.26

 honor

2.24

 honour

2.14

 honorable

2.06

 dishon

1.95

Hon

1.86

 honors

1.77

Activations Density 0.024%

words related to recognition or honor

references to the name "John" and variations of its spelling

No Comments

No Known Activations