INDEX

Explanations

words relating to television broadcasts

oai_token-act-pair · gpt-3.5-turbo

instances of the word "televised" and related terms indicating broadcast events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 weeds

-0.69

 Navigation

-0.69

 Builder

-0.68

 Barbarian

-0.68

 Dock

-0.67

 foliage

-0.66

 blockade

-0.66

 blockers

-0.65

pole

-0.65

 decoration

-0.65

POSITIVE LOGITS

 televised

2.41

 tapes

1.55

iem

1.40

ulz

1.37

 videot

1.13

chio

1.07

 cass

1.06

 taped

1.03

 declass

1.03

 cand

1.01

Activations Density 0.043%

words relating to television broadcasts

instances of the word "televised" and related terms indicating broadcast events

No Comments

No Known Activations