INDEX

Explanations

mentions of talk shows and crimes, especially war crimes

oai_token-act-pair · gpt-3.5-turbo

references to talk shows and legal terms related to crimes and penalties

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

BIT

-1.05

bits

-0.95

bit

-0.89

Bit

-0.87

rez

-0.81

ãĤ¨ãĥ«

-0.81

cloud

-0.81

ãĥĥãĥī

-0.78

bats

-0.77

clair

-0.77

POSITIVE LOGITS

 analogy

0.72

 optimization

0.70

 pamph

0.68

 mentality

0.67

 modernization

0.67

 Atlantis

0.66

jur

0.65

 registry

0.64

 imitation

0.64

 fiasco

0.64

Activations Density 0.385%

mentions of talk shows and crimes, especially war crimes

references to talk shows and legal terms related to crimes and penalties

No Comments

No Known Activations