INDEX

Explanations

dates or time-related phrases preceded by the word "then"

oai_token-act-pair · gpt-3.5-turbo

references to roles or positions held by individuals in a chronological context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 theorem

-0.69

 ticks

-0.68

zsche

-0.65

 shade

-0.64

owl

-0.64

 flex

-0.63

 modification

-0.63

 âĿ

-0.63

 Tune

-0.62

 shading

-0.62

POSITIVE LOGITS

cand

1.34

president

1.21

director

1.08

Senator

1.07

govern

1.05

Soviet

1.05

current

1.05

general

1.04

girlfriend

1.04

Govern

1.03

Activations Density 0.024%

dates or time-related phrases preceded by the word "then"

references to roles or positions held by individuals in a chronological context

No Comments

No Known Activations