INDEX

Explanations

comparisons between different concepts or entities

oai_token-act-pair · gpt-3.5-turbo

comparative phrases that highlight moral or ethical considerations

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

icken

-0.72

eds

-0.70

Bern

-0.67

âĢĲ

-0.66

encers

-0.65

Domain

-0.62

medium

-0.59

hiba

-0.59

bags

-0.58

engine

-0.58

POSITIVE LOGITS

 slapping

0.75

 having

0.73

 brute

0.71

 assass

0.71

 abol

0.70

 anything

0.69

 removing

0.68

 rewriting

0.68

 blasphemy

0.68

 curing

0.68

Activations Density 0.199%

comparisons between different concepts or entities

comparative phrases that highlight moral or ethical considerations

No Comments

No Known Activations