APOLLO: An Optimized Training Approach for Long-form Numerical Reasoning (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods to generate reasoning programs that ignore the differences between facts treated all facts equally, leading to wrong punishment of programs that differed from the ground truth. |
| Approach: | They propose an optimized training framework for long-form numerical reasoning that incorporates a number-aware negative sampling strategy and consistency-based reinforcement learning to increase execution accuracy. |
| Outcome: | The proposed method improves the performance of long-form numerical reasoning on the FinQA and ConvFinQA leaderboards. |
Similar Papers
APOLLO: A Simple Approach for Adaptive Pretraining of Language Models for Logical Reasoning (2023.acl-long)
Copied to clipboard
Soumya Sanyal, Yichong Xu, Shuohang Wang, Ziyi Yang, Reid Pryzant, Wenhao Yu, Chenguang Zhu, Xiang Ren
| Challenge: | Existing methods to improve logical reasoning skills require complex data processing. |
| Approach: | They propose an adaptive pretraining approach to improve logical reasoning over text . they use a subset of Wikipedia sentences for pretraining and a sentence-level classification loss . |
| Outcome: | The proposed model outperforms baselines on LogiQA and ReClor. |
Enhancing Numerical Reasoning with the Guidance of Reliable Reasoning Processes (2024.acl-long)
Copied to clipboard
| Challenge: | Numerical reasoning is an essential ability for NLP systems to handle numeric information. |
| Approach: | They propose a numerical reasoning method that generates reliable reasoning processes by decomposing the answer formula and aim to train models to generate the process with synthesized data. |
| Outcome: | The proposed method improves on all five datasets with an average improvement of 1.8% compared with baselines and gpt-3.5-turbo. |
An Efficient and Precise Training Data Construction Framework for Process-supervised Reward Model in Mathematical Reasoning (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for constructing process supervision training data are costly or suffer from poor quality. |
| Approach: | They propose a framework called EpicPRM which annotates each intermediate reasoning step based on its quantified contribution and uses an adaptive binary search algorithm to enhance annotation precision and efficiency. |
| Outcome: | The proposed framework improves annotation precision and efficiency and can be used to train a high-quality training dataset with 50k annotated intermediate steps. |
FERMAT: An Alternative to Accuracy for Numerical Reasoning (2023.acl-long)
Copied to clipboard
| Challenge: | Existing numerical reasoning models are too weak for downstream tasks like fact-checking . FERMAT evaluates models on number understanding, mathematical operations, and training dependency . |
| Approach: | They propose a multi-view evaluation set for numerical reasoning in English that evaluates models on key numerical reasoning aspects instead of reporting a single score on a whole dataset. |
| Outcome: | FERMAT evaluates models on number understanding, mathematical operations, and training dependency. |
Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) increase test-time computation, often in the form of chain-of-thought (CoT) however, reasoning traces can become unnecessarily long, increasing computation costs without improving accuracy and sometimes even degrading performance. |
| Approach: | They propose a multi-stage efficient reasoning method that combines supervised fine-tuning with reinforcement learning using an adaptive length penalty. |
| Outcome: | The proposed method reduces response length by an average of 28% for 8B models and 40% for 32B models while incurring only minor performance drops of 1.6 and 2.5 points, respectively. |
Self-training Language Models for Arithmetic Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work improves the reasoning capabilities of language models by scaling training data to more diverse or complex collections, but reaching further improvements becomes exceedingly expensive. |
| Approach: | They propose to use implicit feedback to improve models' reasoning capabilities by training from implicit feedback. |
| Outcome: | The proposed model can reach a correct result in +13.9% and +25.9% more cases than previous models, underlining the importance of actuality of self-training feedback. |
Predicate-Guided Generation for Mathematical Reasoning (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experimental results show that Prolog-MATH generates 81.3% solution coverage on Deepseek-V3 . |
| Approach: | They propose a curated corpus to support mathematical reasoning in large language models . they propose supervised fine-tuning followed by GRPO training to address problems that Deepseek-V3 fails to solve. |
| Outcome: | The proposed pipeline achieves 81.3% solution coverage on the Deepseek-V3 training set. |
Select2Reason: Efficient Instruction-Tuning Data Selection for Long-CoT Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large reasoning models exhibit human-like behaviors such as exploration, verification, reflection, and correction. |
| Approach: | They propose a supervised fine-tuning framework for long chain-of-thoughts reasoning . they leverage a difficulty-aware reward model to estimate the learning value of questions . |
| Outcome: | The proposed framework performs fine-tuning on large reasoning models on 10% of the data selected. |
Towards Table-to-Text Generation with Numerical Reasoning (2021.acl-long)
Copied to clipboard
| Challenge: | Recent studies have shown improvement in generating descriptive text from structured data. |
| Approach: | They propose a framework for numerical table-to-text generation based on numerical reasoning . they use a pre-trained model and a copy mechanism to fine-tune the models to produce fluent text . |
| Outcome: | The proposed framework lacks fidelity to the table contents and is based on a pre-trained model and a copy mechanism. |
EpiCaR: Knowing What You Don’t Know Matters for Better Reasoning in LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to improving reasoning abilities of large language models incur a significant calibration cost. |
| Approach: | They propose an epistemic learning problem that integrates reasoning and calibration into an iterative supervised training framework. |
| Outcome: | The proposed method achieves Pareto-superiority over standard baselines in accuracy and calibration. |