Generating academic paper titles requires balancing two key goals: faithfulness to the content and attractiveness to readers. Existing methods often treat title generation as a simple summarization task, producing accurate but unengaging titles. While multi-objective Direct Preference Optimization (DPO) offers a potential solution, our empirical analysis shows that combining conflicting objectives into a single reward is associated with substantial gradient interference during training. To address this problem, drawing inspiration from Pareto dominance in multi-objective optimization, we propose Multi-Objective Dominance Alignment (MODA), a data-centric training strategy tailored for academic title generation. MODA retains preference pairs with strict Pareto dominance, thereby reducing exposure to supervision with opposing reward directions. We implement MODA using dual reward models to independently score both dimensions. Across our experiments, this filtering strategy improves the observed balance between faithfulness and attractiveness, outperforms Human Author baselines, and approaches the performance of frontier models despite using an 8B backbone. Our code is publicly available at https://github.com/realyoume/title_it_right.

Evidence-intensive analytical reports are expected to be fact-dense, quantitatively correct, and supported by figures. Yet one-shot long-form generation with large language models (LLMs) frequently produces fluent but under-supported drafts: core facts are missed, numbers drift, and key visuals are absent, making the report hard to trust. We propose EviReport, an evidence-tracked report-writing workflow that improves reliability by (i) organizing corpus evidence into compact, traceable units and retrieves query-relevant subgraphs into retrieval-ready packages (ii) leveraging a reasoning-focused LLM sketches a high-level plan for full coverage, then a chat-based LLM sharpens it into a detailed hierarchical outline with explicit scope and ordering (iii) rive generation with a facts-first iterative loop: extracting verifiable facts, composing strictly from those facts, then triggering gap-aware append queries to fill missing evidence To evaluate both correctness and completeness, we introduce EviReportBench, a benchmark instantiated on data-rich indicator reports that measures factual accuracy (claim verification), factual coverage (quiz-based evaluation), and visual evidence integration (image recall). Across 8 topics, experiments show that EviReport consistently outperforms strong baselines in factual coverage (2.16×), factual accuracy (+8.9 points), and visual evidence integration (+34 points), approaching the quality of expert-written reports across multiple dimensions.

Large-scale content moderation requires models that can apply complex, evolving policies under limited expert supervision. We present an empirical study of reinforcement learning (RL) for policy-grounded moderation across three production-derived tasks. Using established RL and inference primitives, we develop three integrity-specific designs: a policy-grounded full-trace reward, rationale-marginalized Monte Carlo scoring, and policy-structured reflection. Each design addresses an observed failure mode, including reasoning-length collapse and polarized confidence estimates. Across the three tasks, RL-Only systems trained on 200–836 labeled examples remain competitive with supervised fine-tuning (SFT) systems trained on thousands of examples, corresponding to up to approximately one order of magnitude greater label efficiency at the evaluated operating points. Performance generally improves with additional data and training compute before saturating, although larger rollout counts exhibit variability. The strongest SFT→RL systems attain point-estimate performance comparable to policy-trained full-time employee (FTE) reviewers across the three tasks.