Although large language model agents are increasingly applied to quantitative-finance workflows, their evaluation remains fragmented across isolated tasks, while the financial relevance of benchmark tasks is often overlooked. Yet financial workflows are inherently multi-stage, spanning interdependent tasks such as forecasting, strategy construction, risk management, and trading. Existing platforms typically focus on a single task, and can therefore overstate agent competence and fail to reveal weaknesses in generalization, real-market interaction, and financially meaningful decision-making. We introduce OpenFinGym, a unified gym environment for quantitative-finance agent development that covers forecasting, market generation, real-time trading, and fraud detection under a single execution and verification interface. OpenFinGym additionally provides an automated task-construction pipeline that turns quantitative finance publications into executable task packages; a containerised runtime with a host-side verifier service that supports scalable agent rollouts and prevents runtime train–test leakage; a paper trading engine with a low-latency data-stream design; deferred-resolution support for long-horizon and event-market forecasts; and integration for SFT and RL post-training. Code is available at https://github.com/DeepIntoStreams/OpenFinGym.git

Conformal prediction (CP) is a distribution-free framework for uncertainty quantification that has recently been adapted to large language models (LLMs), providing prediction sets with finite-sample coverage guarantees under exchangeability. Yet for LLMs, nonconformity scores are often induced by an inference pipeline, not just a fixed model, making them depend not only on the data distribution but also on configurable factors such as the prompt template, decoding parameters, and deployment setting. Since such configurations are routinely modified in practice but rarely treated as a source of shift, their impact on CP validity remains poorly understood. We call this configuration shift and study it systematically along three axes: prompt template, decoding temperature, and weight quantization. In a broad empirical study spanning 9 LLMs, 4 datasets, and 4 nonconformity scores, we find that configuration shift consistently erodes CP validity, often driving empirical coverage below the target. By contrast, efficiency is largely preserved: valid prediction sets remain close in size to the i.i.d. baseline. We derive coverage lower bounds that attribute this loss to a discrepancy between calibration and test score distributions, and use their finite-sample plug-in versions as empirical diagnostics of shift severity. We further show that these findings lead to practical mitigations: bound-inspired recalibration is effective with limited test examples, while fragility-aware calibration ensembling recovers much of the lost coverage without test data.

This paper details the system description and approach used by our team for the SMM4H 2020 competition, Task 1. Task 1 targets the automatic classification of tweets that mention medication. We adapted the standard BERT pretrain-then-fine-tune approach to include an intermediate training stage with a biLSTM architecture neural network acting as a further fine-tuning stage. We were inspired by the effectiveness of within-task further pre-training and sentence encoders. We show that this approach works well for a highly imbalanced dataset. In this case, the positive class is only 0.2% of the entire dataset. Our model performed better in both F1 and precision scores compared to the mean score for all participants in the competition and had a competitive recall score.