INDEX

Explanations

mentions of legal actions or authorities seizing items

oai_token-act-pair · gpt-3.5-turbo

phrases associated with claims or allegations, particularly regarding data loss or access issues

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

 honoring

-0.89

 canceled

-0.82

 harbor

-0.79

 honored

-0.76

 someday

-0.75

colored

-0.75

 honors

-0.74

 honor

-0.74

 airplanes

-0.72

 coloring

-0.71

POSITIVE LOGITS

 campaigner

1.41

 campaigners

1.34

 UKIP

1.30

 Parliamentary

1.27

MEP

1.27

 Labour

1.26

 organisation

1.25

 OECD

1.20

 organisations

1.19

NHS

1.16

Activations Density 0.592%

mentions of legal actions or authorities seizing items

phrases associated with claims or allegations, particularly regarding data loss or access issues

No Comments

No Known Activations