Mova · papers by stage
frames → world
Keeping the world of the text: entities, their states, quantities and compiled commonsense. The 9 works of Mova's own list that this stage uses; measured marks a work whose effect Mova measured in its own experiments (5 here). Each title opens our card: the summary, how Mova uses the work, its references and who cites it. What the stage does is on Mova's pipeline page; every stage is on the main list.
- Solving General Arithmetic Word ProblemsmeasuredSubhro Roy, Dan Roth · 2015
What Mova took:
math/src/arith.rs— the core of the arithmetic tree model: a relevance judgement for each number and an operation judgement for each pair of numbers at their lowest common ancestor, both as averaged perceptrons over word… - Unit Dependency Graph and its Application to Arithmetic Word Problem SolvingmeasuredSubhro Roy, Dan Roth · 2017
What Mova took:
global/seeds/shortcuts/units.md: a first-level knowledge "seed" that states the core rule (a "X per Y" quantity is a rate; total = rate × number of units; equal units combine only by addition/subtraction) and links concepts… - COMET-ATOMIC 2020: On Symbolic and Neural Commonsense Knowledge GraphsmeasuredJena D. Hwang et al. · 2020 (arXiv) / 2021 (AAAI)
What Mova took: Mova uses the ATOMIC 2020 data (CC BY 4.0), not the neural COMET model. Small SQL extractions (
global/data/atomic-feelings.sql,global/data/atomic-uses.sql) produce compiled first-level seeds. No knowledge model is trained. - Learning to Reason Deductively: Math Word Problem Solving as Complex Relation ExtractionmeasuredZhanming Jie, Jierui Li, Wei Lu · 2022
What Mova took:
math/src/deduct.rs(math solver v5,mathsolve steps --dag): a transparent reimplementation of the DeductReasoner idea as a graph of subproblems (DAG). The pool holds the numbers from the text plus all intermediate results… - Cutting Through the Noise: Boosting LLM Performance on Math Word ProblemsmeasuredUjjwala Anantheswaran et al. · 2024
What Mova took: Inspired the
--augmentoption of the math solver training in math/src/bin/mathsolve.rs: every training problem gets a copy with a sentence containing a number from another problem inserted just before the question, while the… - Categorization and Representation of Physics Problems by Experts and NovicesMichelene T. H. Chi, Paul J. Feltovich, Robert Glaser · 1981
What Mova took:
world/src/quant.rs— the quantitative world layer for text word problems: each state change in a script is tied to a sentence (@N) and labelled with the principle it represents… - Learning to Solve Arithmetic Word Problems with Verb CategorizationMohammad Javad Hosseini et al. · 2014
What Mova took:
math/src/steps.rs: the math solver's main non-lexical features are verb classes (have, get, give, make, lose, move, compare) in the spirit of ARIS's verb categorisation, extended with Roy & Roth 2018. The model sees "how the… - Learning To Use Formulas To Solve Simple Arithmetic ProblemsArindam Mitra, Chitta Baral · 2016
What Mova took:
global/seeds/shortcuts/units.md— the comparison formula is the source of a first-level link "compare → subtract": questions of the form "how many more / fewer" are explained as a difference, i.e. subtraction. - Mapping to Declarative Knowledge for Word Problem SolvingSubhro Roy, Dan Roth · 2018
What Mova took:
global/seeds/shortcuts/possession.md: a first-level knowledge seed with verb classeshave,get,give,make,lose,move…