Financial language agents are commonly evaluated through static prediction tasks or isolated historical backtests, where an agent’s decisions cannot affect subsequent prices or the behavior of competing strategies. This overlooks a defining property of real markets: interpretations of financial news influence order flow, price formation, and the competitive environment faced by other participants. We introduce FINEVO, a closed-loop evaluation framework that integrates time-stamped financial news, heterogeneous trading agents, a continuous double-auction market, and population-level capital reallocation. FINEVO evaluates language-based agents alongside technical, forecasting, reinforcement-learning, and informed strategies, reporting not only returns but also market concentration, strategic reorganization, regime-dependent standing, and recovery from shocks. Across 128-run suites of controlled market shocks and historical-news-conditioned episodes, we find that language-agent performance is strongly contingent on the surrounding market regime and agent population. Endogenous price formation also causes identical information schedules to produce different outcomes under different competitive compositions. Mechanism ablations further show that innovation is critical for maintaining contestability: removing it increases mean HHI from 0.189 to 0.477 and mean Top-1 capital share from 0.344 to 0.531. These results suggest that financial language agents should be evaluated not only by how accurately they interpret information, but also by how their decisions reshape the markets in which they operate.

Large Language Models (LLMs) exhibit strong mathematical reasoning when trained on high-quality Chain-of-Thought (CoT) that articulates intermediate steps, yet costly CoT curation hinders further progress. While existing remedies such as distillation from stronger LLMs and self-synthesis based on test-time search alleviate this issue, they often suffer from diminishing returns or high computing overhead. In this work, we propose CoTEvol, a genetic evolutionary framework that casts CoT generation as a population-based search over reasoning trajectories. Candidate trajectories are iteratively evolved through reflective global crossover at the trajectory level and local mutation guided by uncertainty at the step level, enabling holistic recombination and fine-grained refinement. Lightweight, task-aware fitness functions are designed to guide the evolutionary process toward accurate and diverse reasoning. Empirically, improves correct-CoT synthesis success by over 30% and enhances structural diversity, with markedly improved efficiency. LLMs trained on these evolutionary CoT data achieve an average gain of 6.6% across eight math benchmarks, outperforming previous distillation and self-synthesis approaches. These results underscore the promise of evolutionary CoT synthesis as a scalable and effective method for mathematical reasoning tasks.

Scientific idea generation requires more than fluent continuation from a problem statement: strong ideas often emerge from inspecting the same problem through multiple informative perspectives. We propose PRISM, an entropy-guided multi-view framework for scientific ideation. Given a research problem, PRISM constructs candidate inspection views, estimates their pairwise orthogonality through hypothesis-transition entropy, selects a maximally orthogonal subset, and conditions a backbone model on the selected views to synthesize a single proposal-style idea. We evaluate PRISM on 500 research problems across 23 scientific domains using three generation backbones: Llama-3.1-8B-Instruct, Qwen3-8B, and GPT-5.4. To assess generated ideas, we use two complementary automated judges—a Galactica-based impact discriminator trained from citation-derived scientific preferences and GPT-5.5 blind five-way novelty ranking—as well as a blind human study with five PhD-level evaluators. Across 7,500 generated ideas, PRISM achieves the best impact score and an 80.2% impact pairwise win rate, a 90.9% novelty pairwise win rate, and the highest mean score (3.91/5) in human evaluation. These results suggest that entropy-guided multi-view exploration provides an effective inference-time mechanism for producing more novel and impactful scientific ideas. Our code is available at https://github.com/jiajunyu1999/PRISM-idea-generation.

Financial decision-making presents unique challenges for language models, requiring them to handle temporally evolving, risk-sensitive, and event-driven contexts. While large language models (LLMs) demonstrate strong general reasoning abilities, they often overlook key behavioral patterns underlying human financial behavior—such as expert reliance under information asymmetry, loss-averse risk adjustment, and temporal adaptation. We propose FinHEAR, a multi-agent framework for Human Expertise and Adaptive Risk-aware reasoning. FinHEAR coordinates multiple LLM-based agents to capture historical trends, interpret current events, and incorporate expert knowledge within a unified, event-aware pipeline. Grounded in behavioral economics, FinHEAR features mechanisms for expert-guided retrieval to reduce information asymmetry, dynamic position sizing to reflect loss aversion, and feedback-driven refinement to enhance temporal consistency. Experiments on a curated real-world financial dataset show that FinHEAR consistently outperforms strong baselines in both trend forecasting and decision-making.

Large language models (LLMs) have advanced general-purpose reasoning, showing strong performance across diverse tasks. However, existing methods often rely on implicit exploration, where the model follows stochastic and unguided reasoning paths—like walking without a map. This leads to unstable reasoning paths, lack of error correction, and limited learning from past experience. To address these issues, we propose a framework that shifts from implicit exploration to structured reasoning through guideline and refinement. First, we extract structured reasoning patterns from successful trajectories and reflective signals from failures. During inference, the model follows these guidelines step-by-step, with refinement applied after each step to correct errors and stabilize the reasoning process. Experiments on the Big-Bench Hard (BBH) benchmark show that our method consistently outperforms strong baselines across diverse reasoning tasks. Analysis reveals that stepwise execution, refinement, and experience-based learning improve stability and generalization. We further explore model collaboration during refinement, offering insights into cross-model interactions. Notably, structured reasoning guided by learned instructions matches or even surpasses knowledge distilled through SFT, highlighting its scalability and effectiveness.