INDEX

Explanations

phrases indicating something has not happened or been done yet

oai_token-act-pair · gpt-3.5-turbo

the phrase "yet to" in various contexts indicating ongoing action or pending decisions

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

rities

-0.79

packs

-0.76

stories

-0.76

 Gems

-0.76

warts

-0.75

gnu

-0.74

equality

-0.72

rats

-0.70

performance

-0.69

current

-0.69

POSITIVE LOGITS

 succumb

0.70

 surpassed

0.70

uden

0.69

 mirac

0.69

 hesitated

0.69

 ascert

0.68

 officially

0.68

 definitively

0.67

 ruled

0.66

 decisively

0.64

Activations Density 0.014%

phrases indicating something has not happened or been done yet

the phrase "yet to" in various contexts indicating ongoing action or pending decisions

No Comments

No Known Activations