INDEX

Explanations

verbs that indicate an action or state of being, often in a negative context

oai_token-act-pair · gpt-3.5-turbo

words indicating the state of being or actions not accomplished

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ortment

-0.70

adder

-0.65

itton

-0.64

Discussion

-0.62

 hopefully

-0.61

rather

-0.60

DragonMagazine

-0.60

MAY

-0.60

 doubtless

-0.60

 assorted

-0.58

POSITIVE LOGITS

 anymore

1.93

nor

1.56

 anything

1.23

any

1.18

 anywhere

1.16

 whatsoever

1.16

 necessarily

1.14

 anybody

1.13

 slightest

1.04

 anyone

1.01

Activations Density 0.881%

verbs that indicate an action or state of being, often in a negative context

words indicating the state of being or actions not accomplished

No Comments

No Known Activations