INDEX

Explanations

terms related to hidden or underground online activities

oai_token-act-pair · gpt-3.5-turbo

references to online platforms and communities

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

 negotiators

-0.80

 Auditor

-0.79

HAEL

-0.74

 Cullen

-0.70

reath

-0.70

enary

-0.69

 conductor

-0.68

Pwr

-0.66

 delegation

-0.65

psey

-0.65

POSITIVE LOGITS

 forums

1.52

 subreddits

1.48

Reddit

1.42

 websites

1.39

 subreddit

1.33

 sites

1.31

forums

1.27

 blogs

1.26

 Reddit

1.26

 Craigslist

1.25

Activations Density 0.457%

terms related to hidden or underground online activities

references to online platforms and communities

No Comments

No Known Activations