INDEX

Explanations

texts with a focus on in-depth exploration or analysis

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

ATT

-0.69

oppable

-0.69

Ĥİ

-0.67

EED

-0.65

haw

-0.64

advertising

-0.63

ãĥ¼ãĥ³

-0.62

 Ready

-0.62

Frames

-0.62

amazon

-0.61

POSITIVE LOGITS

 vein

1.08

 trenches

1.04

dug

1.01

 deep

1.00

 depths

0.97

 deeper

0.95

deep

0.94

 pockets

0.92

 trench

0.90

 deepest

0.89

Activations Density 0.909%

texts with a focus on in-depth exploration or analysis

No Comments

No Known Activations