INDEX

Explanations

comparisons using the phrase "not as" followed by a subjective quality being examined

oai_token-act-pair · gpt-3.5-turbo

comparisons using the word "as."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Leaks

-0.77

PLA

-0.74

WAYS

-0.72

 guiActive

-0.69

é¾

-0.68

DIT

-0.67

raltar

-0.65

 sqor

-0.64

çĶ°

-0.64

å§«

-0.63

POSITIVE LOGITS

 flashy

1.09

 glamorous

1.02

 drastic

0.99

 easily

0.96

 egregious

0.96

bad

0.95

 pronounced

0.95

 readily

0.95

 easy

0.95

 robust

0.94

Activations Density 0.044%

comparisons using the phrase "not as" followed by a subjective quality being examined

comparisons using the word "as."

No Comments

No Known Activations