INDEX

Explanations

specific organizations, research-related terms, and contact information

oai_token-act-pair · gpt-3.5-turbo

references to research funding and institutional affiliations

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

 revenge

-0.95

venge

-0.92

 booze

-0.81

 draining

-0.79

 combust

-0.77

 desperate

-0.75

 worse

-0.75

 swapping

-0.73

 rivals

-0.72

 Furious

-0.72

POSITIVE LOGITS

 Academic

1.02

 Scholarship

0.97

 Publications

0.96

 Faculty

0.94

negie

0.93

 Advisory

0.92

ospons

0.91

 Cosponsors

0.91

uscript

0.89

 scholarly

0.89

Activations Density 0.983%

specific organizations, research-related terms, and contact information

references to research funding and institutional affiliations

No Comments

No Known Activations