INDEX

Explanations

references to being physically high on substances

oai_token-act-pair · gpt-3.5-turbo

references to high levels or abundance in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

issors

-0.69

ÄŁ

-0.66

sbm

-0.60

Wik

-0.60

worker

-0.59

 Folder

-0.59

Factory

-0.58

 principals

-0.58

workers

-0.57

Äĩ

-0.57

POSITIVE LOGITS

(>

0.75

ritic

0.70

antage

0.69

fal

0.66

resolution

0.63

sever

0.63

odic

0.63

ensity

0.62

olesterol

0.62

 achie

0.61

Activations Density 0.145%

references to being physically high on substances

references to high levels or abundance in various contexts

No Comments

No Known Activations