INDEX

Explanations

adjectives expressing lack of harm or danger

oai_token-act-pair · gpt-3.5-turbo

terms related to the concepts of harmlessness and benignity

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

KER

-0.70

GPU

-0.69

lining

-0.66

funding

-0.64

Cla

-0.64

Pain

-0.62

pain

-0.62

ingo

-0.61

lin

-0.61

Connell

-0.60

POSITIVE LOGITS

 harmless

1.04

 innocuous

0.92

 benign

0.85

»Ĵ

0.81

 minded

0.81

alty

0.78

 bystand

0.77

ality

0.76

mate

0.71

ishable

0.69

Activations Density 0.016%

adjectives expressing lack of harm or danger

terms related to the concepts of harmlessness and benignity

No Comments

No Known Activations