INDEX

Explanations

information related to winning or achievements

oai_token-act-pair · gpt-3.5-turbo

instances of the word "won" related to achievements or victories

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

OTOS

-0.69

erity

-0.68

assies

-0.66

angering

-0.62

avia

-0.62

bian

-0.62

 footprint

-0.61

appa

-0.61

repre

-0.60

tools

-0.60

POSITIVE LOGITS

't

1.21

itive

0.82

ners

0.72

ALD

0.71

kish

0.68

æ©

0.68

now

0.68

rar

0.68

Won

0.66

geon

0.64

Activations Density 0.023%

information related to winning or achievements

instances of the word "won" related to achievements or victories

No Comments

No Known Activations