INDEX

Explanations

discern names and personal details of individuals, likely related to news articles

oai_token-act-pair · gpt-3.5-turbo

references to individuals and their characteristics or fates in a narrative context

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Feedback

-0.65

¥µ

-0.62

________________

-0.62

-0.61

 Watching

-0.60

Higher

-0.59

âĻ¦

-0.58

 spons

-0.58

âĶģ

-0.58

inline

-0.58

POSITIVE LOGITS

 died

1.28

 succumbed

1.22

 perished

1.10

 underwent

1.10

 belonged

1.10

 survived

1.09

 resided

1.08

 suffered

1.05

 lived

1.05

 vanished

1.03

Activations Density 0.161%

discern names and personal details of individuals, likely related to news articles

references to individuals and their characteristics or fates in a narrative context

No Comments

No Known Activations