INDEX

Explanations

names related to "Mon-" and images or nudity

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

BW

-0.66

ETF

-0.61

 indent

-0.58

 hindsight

-0.56

 Beir

-0.55

 insign

-0.55

 Flags

-0.54

 dividends

-0.54

 stewards

-0.54

writing

-0.54

POSITIVE LOGITS

theless

0.81

hao

0.70

vre

0.67

opol

0.66

liction

0.66

atari

0.66

rals

0.65

phrine

0.65

 Carlo

0.65

 Xuan

0.65

Activations Density 0.057%

names related to "Mon-" and images or nudity

No Comments

No Known Activations