INDEX

Explanations

terms related to different types of attacks in a combat context (ranged, melee)

oai_token-act-pair · gpt-3.5-turbo

terms related to ranged and melee combat

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

redo

-0.88

mins

-0.84

urated

-0.80

oos

-0.80

agate

-0.77

gro

-0.77

Afee

-0.76

oria

-0.74

src

-0.74

neath

-0.73

POSITIVE LOGITS

 ranged

0.94

 weapons

0.84

 combatants

0.82

 projectiles

0.81

 melee

0.77

 weaponry

0.77

EStreamFrame

0.75

 frontline

0.74

cas

0.73

 gunfire

0.72

Activations Density 0.010%

terms related to different types of attacks in a combat context (ranged, melee)

terms related to ranged and melee combat

No Comments

No Known Activations