INDEX

Explanations

references to specific military bases

oai_token-act-pair · gpt-3.5-turbo

references to military or operational bases

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

 curtains

-0.75

 respons

-0.72

Nev

-0.72

Louis

-0.68

uls

-0.66

 telev

-0.64

 exorc

-0.64

smoking

-0.63

 theolog

-0.61

 tast

-0.61

POSITIVE LOGITS

 Base

3.85

 base

2.58

Base

2.57

base

2.43

 BASE

2.24

 bases

1.98

 Curve

1.10

 Depot

1.00

 Headquarters

1.00

 Core

0.99

Activations Density 0.012%

references to specific military bases

references to military or operational bases

No Comments

No Known Activations