INDEX

Explanations

elements in a structured list

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Aber

-0.67

Ath

-0.62

Ao

-0.61

 whist

-0.60

aneous

-0.58

 Huck

-0.58

 Fury

-0.58

 Outs

-0.58

 Gore

-0.57

 Journalism

-0.57

POSITIVE LOGITS

erv

1.08

ening

0.95

icter

0.90

lists

0.88

 alphabet

0.83

ener

0.83

erve

0.83

icles

0.82

 comprehens

0.81

 newcom

0.81

Activations Density 2.232%

elements in a structured list

No Comments

No Known Activations