Antibodies are essential therapeutic molecules, and their complementarity-determining regions (CDRs) form the primary antigen-recognition interface. Recent protein generative models have demonstrated broad capabilities in biomolecular design, yet post-training strategies for downstream objectives remain limited. Standard denoising training operates on noisy states obtained by perturbing native structures, whereas recursive generation proceeds through model-generated intermediate states. For flexible antibody CDR loops such as CDR-H3, this mismatch can allow backbone deviations to accumulate along the denoising trajectory and compromise antigen-facing loop geometry. We introduce ABOPD, an antibody design framework based on on-policy distillation that leverages privileged native geometry during training to supervise states visited along the model's own denoising trajectories. With this fine-grained structural supervision, ABOPD substantially improves structural recovery on RAbD CDR-H3 generation, reducing RMSD by 0.42 {\AA} (from 2.37 {\AA} to 1.95 {\AA}) and outperforming supervised fine-tuning and offline distillation controls, offering a path to higher-fidelity protein design.
Zhuo Yang, Jiaying He, Jiaqing Xie et al.· 0 citations
Many real‐world scientific and industrial applications require the optimization of expensive black‐box functions. Bayesian optimization (BO) provides an effective framework but often struggles with local optima and lacks interpretability. This paper introduces reasoning BO, a novel framework leveraging reasoning models to guide BO sampling while incorporating multi‐agent systems and knowledge graphs for online knowledge accumulation. We evaluate our approach across 10 diverse tasks, including synthetic functions and complex real‐world chemical optimizations. Reasoning BO progressively refines sampling strategies through real‐time insights and hypothesis evolution, identifying high‐performing regions effectively. In the direct arylation task, our method significantly outperformed traditional BO, increasing yield from 25.20% to 60.07%. Furthermore, we demonstrate that smaller LLMs, after post‐training, can achieve performance comparable to larger counterparts. This framework establishes an intelligent cost‐effective optimization system for scientific discovery, combining LLM reasoning with structured knowledge management.
Zhuo Yang, Daolang Wang, Lingli Ge et al.· Materials Genome Engineering...· 0 citations
Benchmark gains are often treated as evidence of greater LLM capability. Yet the same gain can reflect different changes in model behavior. A model may reach new answers, or produce answers that were already within reach. Aggregate scores do not distinguish these changes question by question. We establish a question-level audit under fixed budgets, temperatures, and answer formats. A question is realized when the default deployment procedure produces the correct answer. A question is reachable when a specified probe finds that answer within a fixed budget. We first test whether inference-time layer routing can expand reachability. Under a matched budget, random routes match or exceed structured search in all 43 model and task settings. Answer-blind procedures retain almost none of this gain, which instead requires access to the correct answer. We then ask why reachable answers sometimes fail to appear. Across six cases spanning 0.5B to 31B, silencing one identified MLP block repairs 68 to 92 percent of a predefined failure set. We next test whether training closes the gap by expanding reachability. In five of six matched evaluations, deployed performance rises while the reachable ceiling remains flat or falls. For DAPO, the deployed score rises by 14.7 points while the reachable ceiling falls by 13.3 points. Across the settings we audit, realization and reachability therefore do not always change together. Claims of capability expansion should report both realized performance and reachability under matched evaluation conditions. Code is available at https://github.com/LiZaiyuan0619/reachability-not-realization
Yanchao Li, Wanhao Liu, Jiaqing Xie et al.· 0 citations
AgentFold is presented, a multi-agent framework that formulates folding-model development as a closed-loop search over executable code variants and improves the best lDDT by 7.5% over independent Codex proposals and outperforms a random-search control.
Mingquan Liu, Jiangyue Chen, Hanqun Cao et al.· 0 citations
MASS learns low-dimensional principal manifold coordinates with a dense autoencoder for coarse semantic grouping, and then performs quality-aware sparse feature coverage within each group using a TopK sparse autoencoder and proposes MASS.
Peng Sun, Yi Yang, Antong Zhang et al.· 0 citations
Data-DPO, a target model-oriented SFT data selection method that consistently outperforms existing data selection baselines under multiple data budgets and stably surpasses full data training performance is proposed.
Peng Sun, Yi Yang, Antong Zhang et al.· 0 citations