INDEX

Explanations

names of cards in a specific game

oai_token-act-pair · gpt-3.5-turbo

references to specific card names and effects in a card game context

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

enegger

-0.85

çīĪ

-0.80

 certific

-0.78

 UNCLASSIFIED

-0.77

 referen

-0.76

 misunder

-0.75

 sitcom

-0.74

 acknow

-0.74

 behavi

-0.74

 Sach

-0.74

POSITIVE LOGITS

blade

1.20

Spell

1.11

romancer

1.06

Blade

1.05

 Grimoire

1.04

 Warlock

1.04

 Blades

1.04

Weapon

1.03

 Vengeance

1.03

Sword

1.03

Activations Density 0.234%

names of cards in a specific game

references to specific card names and effects in a card game context

No Comments

No Known Activations