INDEX

Explanations

mentions of winners or winning in competitive contexts

oai_token-act-pair · gpt-3.5-turbo

references to "winner" in various contexts, particularly in competitive scenarios

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

uli

-0.72

otrop

-0.70

 sulf

-0.69

rir

-0.68

Ku

-0.65

ohyd

-0.62

Gau

-0.62

rients

-0.62

otropic

-0.61

allel

-0.60

POSITIVE LOGITS

 winner

1.04

 winners

0.97

 contestant

0.94

 prizes

0.90

 contender

0.85

 losers

0.84

 loser

0.83

 prize

0.82

Winner

0.80

 contenders

0.79

Activations Density 0.045%

mentions of winners or winning in competitive contexts

references to "winner" in various contexts, particularly in competitive scenarios

No Comments

No Known Activations