Paper-to-code reproduction asks scientific AI agents to turn research papers into executable repositories that preserve the paper’s method, protocol and artifacts. This is difficult because the specification is split: explicit paper content such as algorithms, metrics and artifacts is often lost across long agent trajectories, while implicit details such as framework defaults and conventions inherited from related work are absent from the paper. We introduce ReproAgent a four-stage Prepare–Plan–Generate–Repair pipeline built around a persistent implementation contract with two channels: an implementation-requirement channel that turns paper snippets into code obligations, and a reference-evidence channel that retrieves content and structure evidence from related repositories. Both are bound to work packages, projected into file-level contracts, and consumed across generation and repair. On PaperBench Code-Dev, ReproAgent reaches the highest mean score among same-backbone scaffolds under both Claude-Sonnet-4.5 and Gemini-3-Flash. End-to-end channel ablations and per-paper cases support the contribution of both channels.Code and experimental artifacts are publicly available at https://anonymous.4open.science/r/ReproAgent-E760.

LLM agents can generate paper reproduction code, yet often produce scientifically unfaithful implementations. We define this failure mode as semantic drift, where generated code silently diverges from the paper’s specifications. We introduce SemanticAlign-Bench(SA-Bench), a diagnostic benchmark covering 30 papers from ICLR, ICML and NeurIPS 2025. For each paper, we decompose its specifications into atomic and verifiable implementation claims, which we call Semantic Alignment Units (SAUs) and evaluate repositories along four diagnostic dimensions spanning numerical, methodological, protocol and ordering drift. In total, we construct 1,491 SAUs across five ML domains and evaluate 12 generator configurations (4 models 3 scaffolds). Even the strongest configuration (Claude + PaperCoder) achieves a mean SAU score of only 0.301 out of 1.0, with an overall mean of 0.221 across 360 evaluations. A failure taxonomy reveals that agents attempt most requirements but implement them incorrectly, with implementation mismatch and stubs accounting for the majority of zero-scored claims. Our analysis further indicates that scaffolds optimized for executability provide limited leverage for scientific reproduction; narrowing the gap requires scaffolds that prioritize semantic specification verification. The benchmark, annotations and evaluation pipeline are publicly available.

Code:https://anonymous.4open.science/r/SemanticAlign-Bench-0676

Data:https://anonymous-hf.up.railway.app/a/rrgn430zpfui/

Real-world analysis often requires combining directly queryable structured sources (e.g., SQL and CSV) with evidence preserved only in visual documents, such as scanned reports, invoices, and dashboards. Current analytics agents still focus primarily on structured and textual inputs, leaving this cross-modal setting underexplored. We introduce DataCross, a benchmark and agent framework for heterogeneous, insight-driven data analysis. DataCrossBench contains 200 end-to-end tasks across six fine-grained categories: Fund Analysis, FinTech, Internet Development, Internet Firms, Short-form Drama, and AI Healthcare, grouped into four macro sectors and mixing SQL/CSV/JSON/TXT sources with image-based evidence. The benchmark is built with a human-in-the-loop reverse-synthesis pipeline to preserve cross-source dependency while keeping the target insights verifiable. We further propose DataCrossAgent, which coordinates source-specialized agents through intra-source exploration, source prioritization, cross-source pollination, and Reasoning ReAct (reReAct)-based executable verification. Experiments show that DataCrossBench reveals a difficulty gap for most baselines, especially on vision-dependent Hard tasks, and that DataCrossAgent improves average factuality over vanilla GPT-4o from 0.3291 to 0.4270 (+29.7%) while showing a smaller Easy-to-Hard degradation than several baselines. These results position DataCrossBench as a practical testbed for cross-modal heterogeneous data analysis.

Vision–language models (VLMs) have progressed rapidly, but Tibetan remains largely underserved due to the lack of infrastructure for reproducible training and evaluation. To help address this gap, we introduce FTibSuite, a resource-centric foundation for Tibetan VLM research that provides an end-to-end training-and-evaluation workflow and includes human-verified multimodal annotations, partially filling a long-standing shortage of Tibetan multimodal resources. FTibSuite comprises FTibData, FTibBench, and a reproducible baseline model, FTibVLM, built on Qwen3-VL-8B-Instruct. FTibVLM adopts a three-stage adaptation pipeline consisting of Tibetan continual pretraining, image–text alignment, and multimodal instruction tuning. For systematic evaluation, FTibBench adapts five established multimodal benchmarks to Tibetan and offers a reproducible evaluation protocol to support consistent comparisons across models. Specifically, FTibBench includes Tibetan versions of MMBench, MME, POPE, BinaryVQA, and COREVQA. Experiments on FTibBench demonstrate that FTibVLM consistently improves Tibetan multimodal performance. For instance, FTibVLM attains 76.01 accuracy on BinaryVQA, indicating that Tibetan performance can be competitive with high-resource settings on this diagnostic task. We also observe substantial gains on other benchmarks, including an improvement on MMBench (dev) from 42.97 to 67.78 and an increase in POPE-random accuracy from 47.53 to 80.56, underscoring the practical value of the proposed workflow and resources.

Extending large language models (LLMs) to low-resource languages often incurs an “align- ment tax”: improvements in the target lan- guage come at the cost of catastrophic forget- ting in general capabilities. We argue that this trade-off arises from the rigidity of supervised fine-tuning (SFT), which enforces token-level surface imitation on narrow and biased data distributions. To address this limitation, we propose a semantic-space alignment paradigm powered by Group Relative Policy Optimiza- tion (GRPO), where the model is optimized us- ing embedding-level semantic rewards rather than likelihood maximization. This objective encourages meaning preservation through flex- ible realizations, enabling controlled updates that reduce destructive interference with pre- trained knowledge. We evaluate our approach on Tibetan–Chinese machine translation and Ti- betan headline generation. Experiments show that our method acquires low-resource capa- bilities while markedly mitigating alignment tax, preserving general competence more effec- tively than SFT. Despite producing less rigid surface overlap, semantic RL yields higher se- mantic quality and preference in open-ended generation, and few-shot transfer results indi- cate that it learns more transferable and ro- bust representations under limited supervision. Overall, our study demonstrates that reinforce- ment learning with semantic rewards provides a safer and more reliable pathway for inclusive low-resource language expansion.

Low-resource target-language generation is often limited by scarce parallel data, while high-resource source-language monolingual data is abundant but difficult to use with standard supervised fine-tuning. We propose Source-Grounded Semantic Reinforcement Learning (SG-SRL), a resource-utilization framework that converts source-language monolingual data into cross-lingual semantic supervision. SG-SRL performs reference-free reinforcement learning (RL) on source-language data with a cross-lingual semantic reward, followed by lightweight supervised recovery. Although direct reward optimization induces verbosity-based reward hacking, recovery restores fluency, conciseness, and task format without erasing the learned semantic alignment. On Chinese-to-Thai generation, SG-SRL outperforms cold-start SFT, source-only continued pretraining, and synthetic-data SFT; human preference and training-independent metrics corroborate the gains. Recovery ablations further show that semantic retention and form restoration are jointly controlled by recovery data and intensity. An in-domain Chinese-to-Tibetan experiment shows that the reward interface can use a contrastively trained encoder when a capable multilingual reranker is unavailable, while also delimiting the framework’s dependence on reward-model discrimination.

Minority languages in China, such as Tibetan, Uyghur, and Traditional Mongolian, face significant challenges due to their unique writing systems, which differ from international standards. This discrepancy has led to a severe lack of relevant corpora, particularly for supervised tasks like headline generation. To address this gap, we introduce a novel dataset, Chinese Minority Headline Generation (CMHG), which includes 100,000 entries for Tibetan, and 50,000 entries each for Uyghur and Mongolian, specifically curated for headline generation tasks. Additionally, we propose a high-quality test set annotated by native speakers, designed to serve as a benchmark for future research in this domain. We hope this dataset will become a valuable resource for advancing headline generation in Chinese minority languages and contribute to the development of related benchmarks.

While multilingual language models like XLM-R have advanced multilingualism in NLP, they still perform poorly in extremely low-resource languages. This situation is exacerbated by the fact that modern LLMs such as LLaMA and Qwen support far fewer languages than XLM-R, making text generation models non-existent for many languages in the world. To tackle this challenge, we propose a novel framework for adapting multilingual encoders to text generation in extremely low-resource languages. By reusing the weights between the encoder and the decoder, our framework allows the model to leverage the learned semantic space of the encoder, enabling efficient learning and effective generalization in low-resource languages. Applying this framework to four Chinese minority languages, we present XLM-SWCM, and demonstrate its superior performance on various downstream tasks even when compared with much larger models.