INDEX

Explanations

question marks

oai_token-act-pair · gpt-3.5-turbo

questions posed to the audience

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

bidden

-0.72

 domest

-0.69

 recovering

-0.69

 nurs

-0.66

 beginning

-0.64

 acquisitions

-0.64

undai

-0.63

 initialization

-0.63

 fraternity

-0.61

 manif

-0.61

POSITIVE LOGITS

Let

1.32

 Leave

1.32

 Tell

1.28

 Discuss

1.26

 Would

1.26

 Share

1.22

 Comment

1.20

 Feel

1.20

 Submit

1.20

 Vote

1.19

Activations Density 0.100%

question marks

questions posed to the audience

No Comments

No Known Activations