arXiv:2603. 04852v2 Announce Type: replace Abstract: Multi-step theorem prediction is a central challenge in geometry problem solving.
arXiv:2606. 27926v1 Announce Type: new Abstract: Geometry Problem Solving have increasingly adopt the neuro-symbolic paradigm, combining neural intuition with symbolic rigor.
arXiv:2608.30156v1 Announce Type: new Abstract: Plane geometry problem (PGP) solving has become a critical benchmark for multimodal reasoning because it requires accurate visual perception and precis...
G-ReAct is a reasoning framework that frames deep search as state evolution over a fixed-topology query graph, enabling explicit tracking of search progress and constraint preservation. It generates high-quality trajectories for fine-tuning and provides structured guidance during inference without extra fine-tuning. Experiments show that with only 1.9K generated trajectories, a Qwen3 model achieves strong accuracy on BrowseComp-ZH and XBench, outperforming larger open-source baselines, and consistently improves existing LLMs on deep-search tasks.
arXiv:2606. 24965v1 Announce Type: cross Abstract: Reasoning about relational structures remains a significant challenge for neural models, particularly when they must systematically apply learned knowledge to problem instances that are harder than those seen in training.
Hermes introduces a family of harnesses that give models control over how they allocate and reuse context windows during inference, a capability termed contextual reasoning. The accompanying Hermes‑Learn framework trains models in two stages to develop these decision‑making skills, enabling them to scale performance with additional compute at test time. Experiments show that while large models naturally benefit, smaller open‑source models can close the performance gap through this training, with gains generalizing across benchmarks, extrapolating beyond trained compute, and transferring to other scaling methods.
arXiv:2607. 14114v1 Announce Type: cross Abstract: Graph learning under distribution shift presents a persistent challenge, where models adapt to new graphs with limited or even no supervision.
arXiv:2604. 11912v2 Announce Type: replace-cross Abstract: While next-token prediction (NTP) has been the standard objective for training language models, it often struggles to capture global structure in reasoning tasks.
The paper introduces TTL‑SR, a geometry‑aware Test‑Time Learning framework designed to improve quantitative spatial reasoning in visual‑language models. By augmenting queries with geometrically coupled auxiliary prompts, filtering unreliable predictions, and updating models with a geometry‑aware multi‑objective loss on unlabeled test data, TTL‑SR adapts models to new domains without additional 3D supervision. Experiments show substantial accuracy gains on the Q‑Spatial‑ScanNet dataset for two state‑of‑the‑art VLMs.
ReactBench is a benchmark designed to evaluate the structural reasoning abilities of multimodal large language models (MLLMs) using chemical reaction diagrams. The dataset contains 1,618 expert‑annotated question‑answer pairs that test reasoning across four hierarchical task dimensions, from simple endpoint counting to complex topological analysis. Evaluation of 24 MLLMs shows a performance gap of more than 30% between anchor‑based tasks and holistic structural reasoning tasks, indicating that current models struggle with reasoning over branching, converging, and cyclic structures.
arXiv:2601.08747v3 Announce Type: replace-cross Abstract: Current context augmentation methods, such as retrieval-augmented generation, play a crucial role in bridging a model's internal knowledge bo...
arXiv:2509. 04027v4 Announce Type: replace Abstract: Test-time scaling, primarily manifested through multi-step Chain-of-Thought (CoT) reasoning via Reinforcement Learning (RL), has emerged as a pivotal paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs).
The paper introduces FactoSR, a factorized reinforcement learning framework designed to improve spatial reasoning in Vision‑Language Models by addressing a dimensional mismatch between 2D visual inputs and the 3D+temporal nature of the physical world. FactoSR decomposes the reasoning task into three orthogonal geometric sub‑objectives—planar correspondence (XY), depth consistency (Z), and temporal reversibility (T)—and optimizes these constraints within a unified policy learning mechanism. Experiments on multi‑view and video benchmarks show that this decomposition yields significant performance gains, achieving a 5.9% improvement on VSI‑Bench and 4.5% on All‑Angles‑Bench.