tensorkelechi/arxiv-cstext · Datasets at Hugging Face

text stringlengths 0 89.3k
ing batch we propose two strategies to allocate the
budget over different domains
Naive Balancing Since our goal is to mitigate
the bias towards head domains our first balanc
ing policy is to use naive balancing which selects
the same number of inputs for each domain in the
training batch Formally the number of samples
for each domain in the small training batch isB
Kl
where lis the number of domains in the dataset
Adaptive Balancing One of our staged learn
ing frameworks key features is utilizing the fine
tuned student model to actively select representa
tive inputs from wellrepresented domains known
as head domains However employing a naive
balancing policy typically results in the dispropor
tionate allocation of the training budget to datafrom underrepresented domains or tail domains
This training batch may lead the finetuned student
model to struggle to select truly effective examples
from the head domains particularly in the initial
stages Such selections are crucial for the model to
learn effectively from these domains To address
this we implement an adaptive balancing policy
This policy starts by constructing the training batch
with a distribution akin to random selection thus
primarily focusing on head data in the early stages
to warm up the model As the process advances
the policy gradually shifts towards a more balanced
distribution by the final stage ensuring comprehen
sive learning across both head and tail domains
Formally the number of examples for each do
main is the weighted average between the num
bers for random selection and the numbers for
naive balancing For stage i domain d we se
lectnd
NB
KKi
KB
Kli
Kexamples for domain
dto build the training batch for adaptive balanc
ing where Nandndare the total number and the
domain size in the original data D
Then domains are naturally categorized based
on whether the number of required samples per
domain exceeds the available samples in the pool
Domains requiring more samples than available are
designated as head domains for that particular
stage while those with fewer required examples
than available are categorized as tail domains
For tail domains where there are insufficient
samples in the dataset D we rely on the teacher
model to generate both the samples and their cor
responding rationales detailed in Section 34 In
contrast for head domains which have a sufficient
number of samples available to meet the demands
of the training batch we utilize the finetuned stu
dent model to actively select the most representa
tive samples as discussed in Section 35
It is important to note that the classification of
domains as head or tail can vary across different
stages of the training process depending on the
evolving needs and data availability
34 Teacher Data Augmentation
Motivated by the effectiveness of synthetic dataset
generated by blackbox LLMs OpenAI 2023 Rad
ford et al 2019 Zhou et al 2024b we utilize the
teacher LLMs to generate synthetic samples and
corresponding annotations to upsample data for tail
domains To save the annotation budget we require
the teacher model to compose the sample and thecorresponding rationales at the same time
Suppose that we need msynthetic examples for
domain ato satisfy the training batch requirement
Given an instruction following prompt Pc com
posed of three demonstrations from domain a and
teacher model Mt we employ stochastic temper
ature sampling with a fixed temperature and re
peat the process mtimes with generated samples
ˆxa1ˆxamand rationales ˆya1ˆyam
ˆxaiˆyaiMtPc afori 1 m
Then we add the generated samples and ratio
nales to the training batch and combine with the ex
tracted samples from D We present two examples
of synthetic inputs and rationales from the teacher
model in Table 9 in Appendix B The case study
suggests the effectiveness of the teacher model in
generating tail examples
35 Student Active Selection
For head domains our strategy involves actively se
lecting instances from the original dataset to meet
the numeric requirements of the balancing policy
We aim to mitigate information loss from data
downsampling through this active data acquisition
The objective is to identify the most challenging or
uncertain instances for the student model thereby
optimizing its learning trajectory
To quantify instance uncertainty we adapt the
Instruction Following Difficulty IFD metric orig
inally proposed by Li et al 2024ab The IFD
scores are used to measure a training instances

ing batch we propose two strategies to allocate the

budget over different domains

Naive Balancing Since our goal is to mitigate

the bias towards head domains our first balanc

ing policy is to use naive balancing which selects

the same number of inputs for each domain in the

training batch Formally the number of samples

for each domain in the small training batch isB

Kl

where lis the number of domains in the dataset

Adaptive Balancing One of our staged learn

ing frameworks key features is utilizing the fine

tuned student model to actively select representa

tive inputs from wellrepresented domains known

as head domains However employing a naive

balancing policy typically results in the dispropor

tionate allocation of the training budget to datafrom underrepresented domains or tail domains

This training batch may lead the finetuned student

model to struggle to select truly effective examples

from the head domains particularly in the initial

stages Such selections are crucial for the model to

learn effectively from these domains To address

this we implement an adaptive balancing policy

This policy starts by constructing the training batch

with a distribution akin to random selection thus

primarily focusing on head data in the early stages

to warm up the model As the process advances

the policy gradually shifts towards a more balanced

distribution by the final stage ensuring comprehen

sive learning across both head and tail domains

Formally the number of examples for each do

main is the weighted average between the num

bers for random selection and the numbers for

naive balancing For stage i domain d we se

lectnd

NB

KKi

KB

Kli

Kexamples for domain

dto build the training batch for adaptive balanc

ing where Nandndare the total number and the

domain size in the original data D

Then domains are naturally categorized based

on whether the number of required samples per

domain exceeds the available samples in the pool

Domains requiring more samples than available are

designated as head domains for that particular

stage while those with fewer required examples

than available are categorized as tail domains

For tail domains where there are insufficient

samples in the dataset D we rely on the teacher

model to generate both the samples and their cor

responding rationales detailed in Section 34 In

contrast for head domains which have a sufficient

number of samples available to meet the demands

of the training batch we utilize the finetuned stu

dent model to actively select the most representa

tive samples as discussed in Section 35

It is important to note that the classification of

domains as head or tail can vary across different

stages of the training process depending on the

evolving needs and data availability

34 Teacher Data Augmentation

Motivated by the effectiveness of synthetic dataset

generated by blackbox LLMs OpenAI 2023 Rad

ford et al 2019 Zhou et al 2024b we utilize the

teacher LLMs to generate synthetic samples and

corresponding annotations to upsample data for tail

domains To save the annotation budget we require

the teacher model to compose the sample and thecorresponding rationales at the same time

Suppose that we need msynthetic examples for

domain ato satisfy the training batch requirement

Given an instruction following prompt Pc com

posed of three demonstrations from domain a and

teacher model Mt we employ stochastic temper

ature sampling with a fixed temperature and re

peat the process mtimes with generated samples

ˆxa1ˆxamand rationales ˆya1ˆyam

ˆxaiˆyaiMtPc afori 1 m

Then we add the generated samples and ratio

nales to the training batch and combine with the ex

tracted samples from D We present two examples

of synthetic inputs and rationales from the teacher

model in Table 9 in Appendix B The case study

suggests the effectiveness of the teacher model in

generating tail examples

35 Student Active Selection

For head domains our strategy involves actively se

lecting instances from the original dataset to meet

the numeric requirements of the balancing policy

We aim to mitigate information loss from data

downsampling through this active data acquisition

The objective is to identify the most challenging or

uncertain instances for the student model thereby

optimizing its learning trajectory

To quantify instance uncertainty we adapt the

Instruction Following Difficulty IFD metric orig

inally proposed by Li et al 2024ab The IFD

scores are used to measure a training instances