INDEX

Explanations

The main thing this neuron does is find references to hierarchical levels or categories

oai_token-act-pair · gpt-3.5-turbo

references to different tiers or levels in a ranking or classification system

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

okia

-0.80

mercial

-0.77

ãĤŃ

-0.71

hur

-0.69

issance

-0.68

natureconservancy

-0.67

 Evening

-0.64

 Predators

-0.63

romeda

-0.63

vertisements

-0.62

POSITIVE LOGITS

 tier

0.98

 tiers

0.91

uple

0.89

Tier

0.75

 Canter

0.71

ney

0.69

 bracket

0.67

 Tier

0.67

ewater

0.67

robe

0.65

Activations Density 0.015%

The main thing this neuron does is find references to hierarchical levels or categories

references to different tiers or levels in a ranking or classification system

No Comments

No Known Activations