INDEX

Explanations

a pattern formed by specific characters and symbols repeated multiple times

oai_token-act-pair · gpt-3.5-turbo

repeated punctuation or symbol characters

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

gra

-0.64

 urgently

-0.64

 reinvent

-0.62

 Gutenberg

-0.61

 outsourcing

-0.59

ÄŁ

-0.59

 plur

-0.59

 bonded

-0.57

 Guatem

-0.57

 reconc

-0.57

POSITIVE LOGITS

 attRot

0.82

"}

0.77

phrine

0.67

 Carbuncle

0.66

enko

0.66

}\

0.65

Kind

0.65

))))

0.65

 respectively

0.64

genic

0.63

Activations Density 0.172%

a pattern formed by specific characters and symbols repeated multiple times

repeated punctuation or symbol characters

No Comments

No Known Activations