INDEX

Explanations

positive and negative adjectives or phrases related to performance or quality

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hops

-1.05

lees

-1.02

events

-0.99

units

-0.95

casts

-0.95

 Rooms

-0.93

tags

-0.90

iddles

-0.88

ometers

-0.88

tracks

-0.87

POSITIVE LOGITS

 piece

1.08

 swath

1.06

 chunk

1.04

 dose

1.03

 relationship

1.01

 assortment

1.00

 version

1.00

 statement

0.99

 subset

0.99

 standpoint

0.99

Activations Density 2.667%

positive and negative adjectives or phrases related to performance or quality

No Comments

No Known Activations