The field of artificial intelligence is witnessing significant developments in the capabilities of large language models (LLMs) to tackle complex reasoning and problem-solving tasks. Recent research has focused on enhancing the planning and reasoning capabilities of LLMs, enabling them to better address multi-step tasks and provide more coherent and diverse solutions. Notable advancements include the integration of symbolic reasoning frameworks, the development of novel planning paradigms, and the creation of benchmarks to evaluate the performance of LLMs in various domains. Some studies have also explored the limitations of current LLMs, including their reliance on procedural memory and the need for more effective integration of domain knowledge. Overall, these developments are pushing the boundaries of what LLMs can achieve and paving the way for more sophisticated and robust AI systems. Noteworthy papers include SymPlanner, which introduces a novel framework for deliberate planning in LLMs, and FormalMATH, which presents a large-scale benchmark for evaluating the formal mathematical reasoning capabilities of LLMs. Additionally, papers like HyperTree Planning and Recursive Decomposition with Dependencies have proposed innovative approaches to enhance the reasoning capabilities of LLMs.
Advancements in Large Language Models for Complex Reasoning and Problem-Solving
Sources
SymPlanner: Deliberate Planning in Language Models with Symbolic Representation
Understanding LLM Scientific Reasoning through Promptings and Model's Explanation on the Answers
MemEngine: A Unified and Modular Library for Developing Advanced Memory of LLM-based Agents
VECSR: Virtually Embodied Common Sense Reasoning System
HyperTree Planning: Enhancing LLM Reasoning via Hierarchical Thinking
FormalMATH: Benchmarking Formal Mathematical Reasoning of Large Language Models
Recursive Decomposition with Dependencies for Generic Divide-and-Conquer Reasoning
CombiBench: Benchmarking LLM Capability for Combinatorial Mathematics
Knowledge Augmented Complex Problem Solving with Large Language Models: A Survey
Procedural Memory Is Not All You Need: Bridging Cognitive Gaps in LLM-Based Agents
Dragonfly: a modular deep reinforcement learning library
Beyond Theorem Proving: Formulation, Framework and Benchmark for Formal Problem-Solving
Proceedings The 13th International Workshop on Theorem proving components for Educational software
Investigating the Impact and Student Perceptions of Guided Parsons Problems for Learning Logic with Subgoals
The Promise and Limits of LLMs in Constructing Proofs and Hints for Logic Problems in Intelligent Tutoring Systems
MARK: Memory Augmented Refinement of Knowledge