INDEX

Explanations

information related to different individuals and their professional affiliations

oai_token-act-pair · gpt-3.5-turbo

titles, roles, and affiliations of individuals in professional contexts

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

venge

-0.78

Downloadha

-0.70

huh

-0.69

 looted

-0.67

attacks

-0.65

ptoms

-0.65

doms

-0.64

 undercut

-0.63

 revenge

-0.63

bath

-0.63

POSITIVE LOGITS

 trustee

1.24

 honorary

1.20

 advisor

1.17

 Advisory

1.15

ospons

1.15

 Chair

1.15

 lecturer

1.12

 advisory

1.11

 speaker

1.10

 contributor

1.10

Activations Density 0.201%

information related to different individuals and their professional affiliations

titles, roles, and affiliations of individuals in professional contexts

No Comments

No Known Activations