INDEX

Explanations

terms related to official or confidential positions within organizations

oai_token-act-pair · gpt-3.5-turbo

terms related to secretive or classified information

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 neglect

-0.64

ibr

-0.61

assert

-0.61

 Feldman

-0.60

 negligent

-0.60

ICT

-0.60

 pron

-0.59

Paw

-0.59

 bats

-0.58

 impaired

-0.58

POSITIVE LOGITS

artment

0.98

artments

0.97

ary

0.88

lain

0.87

aries

0.83

arily

0.82

 compartment

0.79

roleum

0.79

oslov

0.79

arity

0.78

Activations Density 0.076%

terms related to official or confidential positions within organizations

terms related to secretive or classified information

No Comments

No Known Activations