LLMs increasingly use external tools and stateful environments for complex, multi-step tasks. Further improving their tool-use capabilities, however, remains bottlenecked by the lack of scalable stateful environments and verified multi-turn trajectories matched to the target LLM’s current capability boundary. We observe that agent skills, packages pairing workflow specifications with executable scripts, provide scalable seed material for tool-use environment synthesis. This motivates Grafting, a skill-to-environment synthesis loop in which coding-agent harnesses perform open-ended construction under deterministic validation contracts, producing self-contained executable environments that are refined using trajectory-level verdicts. In an initial end-to-end run, Grafting produces 500 environments and approximately 12K verified multi-turn trajectories. We train on these trajectories with Fathoming, an iterative GRPO procedure that selects verified trajectories near the target model’s capability boundary. Training a 4B-parameter instruction-tuned model with two rounds of Fathoming on the trajectory pool produces Fathom-4B and yields strong out-of-distribution gains: BFCLv4 multi-turn accuracy rises from 22.0 to 39.7 (+17.7 points), and ComplexFuncBench per-call accuracy rises from 49.6 to 61.6 (+12.0 points). Using the same environment-synthesis pipeline, we further construct Grove-5K, a collection of 5,101 executable environments. We release Grafting, Grove-500, Grove-5K, and Fathom-12K.

While reinforcement learning with verifiable rewards (RLVR) has advanced multimodal reasoning, coarse outcome rewards offer limited guidance for assigning credit to visual evidence. Analyzing same-prompt rollout groups, we observe that different outcomes exhibit distinct temporal patterns of measured visual dependency across reasoning segments. We therefore propose **VIGOR** (**Vi**sual **G**rounding **O**utcome-contrastive **R**einforcement), a policy optimization method that converts these group-wise contrasts into fine-grained supervision. VIGOR constructs outcome-conditioned prototypes from temporal visual fingerprints to reweight trajectories, uses segment-level contrasts to prioritize token gradients, and allocates the visual budget according to group difficulty. It requires no additional annotations or reward models and introduces no inference-time overhead. Across eight multimodal reasoning benchmarks, VIGOR achieves 57.5% average accuracy with Qwen2.5-VL-7B, improving over VPPO by 1.6 points. Controlled experiments further validate VIGOR on Qwen2.5-VL-3B and InternVL2.5-4B; on the latter, it outperforms VPPO on all eight tasks by 0.96 points on average. Our code is publicly available at https://github.com/beck83879-bit/VIGOR.

Self-reflection for Large LanguageModels (LLMs) has gained significant attention. Existing approaches involve models iterating and improving their previous responses based on LLMs’ internal reflection ability or external feedback. However, recent research has raised doubts about whether intrinsic self-correction without external feedback may even degrade performance. Based on our empirical evidence, we find that current static reflection methods may lead to redundant, drift, and stubborn issues. To mitigate this, we introduce Instruct-of-Reflection (IoRT), a novel and general reflection framework that leverages dynamic-meta instruction to enhance the iterative reflection capability of LLMs. Specifically, we propose the instructor driven by the meta-thoughts and self-consistency classifier, generates various instructions, including refresh, stop, and select, to guide the next reflection iteration. Our experiments demonstrate that IoRT achieves an average improvement of 10.1% over established baselines in mathematical and commonsense reasoning tasks, highlighting its efficacy and applicability. Our code is available at https://github.com/llp635/IoRT.

The emergence of personalized generation has made it possible to create texts or images that meet the unique needs of users. Recent advances mainly focus on style or scene transfer based on given keywords. However, in e-commerce and recommender systems, it is almost an untouched area to explore user historical interactions, automatically mine user interests with semantic associations, and create item representations that closely align with user individual interests.In this paper, we propose a brand new framework called Interest-Augmented Multimodal Generator (I-AM-G). The framework first extracts tags from the multimodal information of items that the user has interacted with, and the most frequently occurred ones are extracted to rewrite the text description of the item. Then, the framework uses a decoupled text-to-text and image-to-image retriever to search for the top-K similar item text and image embeddings from the item pool. Finally, the Attention module for user interests fuses the retrieved information in a cross-modal manner and further guides the personalized generation process in collaboration with the rewritten text.We conducted extensive and comprehensive experiments to demonstrate that our framework can effectively generate results aligned with user preferences, which potentially provides a new paradigm of Rewrite and Retrieve for personalized generation.