INDEX

Explanations

adverbs ending in -ingly

oai_token-act-pair · gpt-3.5-turbo

phrases and structures related to deception and pretense

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ourses

-0.85

otiation

-0.81

ests

-0.78

cox

-0.78

rer

-0.75

atl

-0.74

mentioned

-0.73

cies

-0.72

olphin

-0.72

summary

-0.71

POSITIVE LOGITS

 invincible

0.85

 unbeat

0.84

 kindred

0.81

 unstoppable

0.79

 innocuous

0.76

 benign

0.75

 harmless

0.74

 resemblance

0.72

 spurious

0.71

 immune

0.70

Activations Density 0.521%

adverbs ending in -ingly

phrases and structures related to deception and pretense

No Comments

No Known Activations