INDEX

Explanations

terms indicating evaluation or judgement

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate something is regarded or viewed in a certain way

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hire

-0.86

dayName

-0.82

inas

-0.76

ionage

-0.71

LET

-0.65

 bombard

-0.65

aptic

-0.65

mouth

-0.65

erers

-0.64

vous

-0.64

POSITIVE LOGITS

phas

0.90

 synonymous

0.90

 taboo

0.84

 worthy

0.83

 kosher

0.83

 canon

0.82

 infall

0.77

 unworthy

0.75

 harmless

0.74

 sacred

0.74

Activations Density 0.039%

terms indicating evaluation or judgement

phrases that indicate something is regarded or viewed in a certain way

No Comments

No Known Activations