INDEX

Explanations

descriptive words and phrases related to performance and behavior

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

lees

-1.12

hops

-1.07

casts

-1.05

units

-1.02

events

-1.00

ernels

-0.97

rooms

-0.96

 Rooms

-0.95

ometers

-0.95

stones

-0.95

POSITIVE LOGITS

 piece

1.11

 standpoint

1.10

 statement

1.07

 swath

1.06

 proposition

1.03

 regimen

1.03

 version

1.03

 chunk

1.02

 dose

1.01

 solution

1.01

Activations Density 4.091%

descriptive words and phrases related to performance and behavior

No Comments

No Known Activations