INDEX

Explanations

proper nouns of people, specifically focusing on descriptions of their actions or roles

oai_token-act-pair · gpt-3.5-turbo

instances of the verb "has" indicating possession or involvement

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Apart

-0.82

Jou

-0.71

 Feminist

-0.71

 Kant

-0.68

 claimants

-0.68

 Capitalism

-0.66

microsoft

-0.65

ieg

-0.65

arity

-0.64

 reviewer

-0.64

POSITIVE LOGITS

 coached

0.83

Coach

0.83

 been

0.80

 teammates

0.79

been

0.79

cussion

0.77

 postseason

0.77

 Heisman

0.76

SIGN

0.76

 concussion

0.75

Activations Density 0.250%

proper nouns of people, specifically focusing on descriptions of their actions or roles

instances of the verb "has" indicating possession or involvement

No Comments

No Known Activations