INDEX

Explanations

words containing the substring "ac" with varying activation strengths

oai_token-act-pair · gpt-3.5-turbo

prefixes that indicate action or occurrence

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

SHIP

-0.80

 shorth

-0.77

 Piper

-0.76

Spa

-0.74

è¦ļéĨĴ

-0.73

 Shots

-0.72

©¶æ¥µ

-0.71

 Painter

-0.70

 Christensen

-0.70

POW

-0.68

POSITIVE LOGITS

ception

1.20

prise

1.05

istant

1.04

ertain

1.01

pect

1.00

mosp

0.97

vance

0.95

ploy

0.94

ighty

0.93

usterity

0.93

Activations Density 0.111%

words containing the substring "ac" with varying activation strengths

prefixes that indicate action or occurrence

No Comments

No Known Activations