INDEX

Explanations

phrases indicating a comparison between different options or sides of an argument

oai_token-act-pair · gpt-3.5-turbo

phrases contrasting two sides of an argument or situation

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

MJ

-0.67

brook

-0.67

DERR

-0.64

 laun

-0.63

nance

-0.63

MSN

-0.62

ciating

-0.62

é¾

-0.62

rations

-0.61

renches

-0.60

POSITIVE LOGITS

 accuser

0.84

 accus

0.62

 accusing

0.62

uthor

0.61

 alone

0.61

 Manafort

0.59

 applaud

0.58

lihood

0.58

 Reps

0.57

icka

0.57

Activations Density 0.035%

phrases indicating a comparison between different options or sides of an argument

phrases contrasting two sides of an argument or situation

No Comments

No Known Activations