INDEX

Explanations

names, especially the name "Joseph"

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

APD

-0.72

 GOODMAN

-0.71

 "$:/

-0.70

hips

-0.69

atron

-0.68

ricted

-0.65

idad

-0.63

raints

-0.63

 warr

-0.62

ADRA

-0.60

POSITIVE LOGITS

smanship

1.04

uth

0.80

hawks

0.77

fires

0.77

fters

0.76

bard

0.72

tein

0.71

fruit

0.71

fire

0.70

ry

0.70

Activations Density 1.549%

names, especially the name "Joseph"

No Comments

No Known Activations