Enduo Zhao

Enduo Zhao, Murilo M. Marinho, Kanako Harada

Robotic assistance for experimental manipulation in the life sciences is expected to enable precise manipulation of valuable samples, regardless of the skill of the scientist. Experimental specimens in the life sciences are subject to individual variability and deformation, and therefore require autonomous robotic control. As an example, we are studying the installation of a cranial window in a mouse. This operation requires the removal of the skull, which is approximately 300 um thick, to cut it into a circular shape 8 mm in diameter, but the shape of the mouse skull varies depending on the strain of mouse, sex and week of age. The thickness of the skull is not uniform, with some areas being thin and others thicker. It is also difficult to ensure that the skulls of the mice are kept in the same position for each operation. It is not realistically possible to measure all these features and pre-program a robotic trajectory for individual mice. The paper therefore proposes an autonomous robotic drilling method. The proposed method consists of drilling trajectory planning and image-based task completion level recognition. The trajectory planning adjusts the z-position of the drill according to the task completion level at each discrete point, and forms the 3D drilling path via constrained cubic spline interpolation while avoiding overshoot. The task completion level recognition uses a DSSD-inspired deep learning model to estimate the task completion level of each discrete point. Since an egg has similar characteristics to a mouse skull in terms of shape, thickness and mechanical properties, removing the egg shell without damaging the membrane underneath was chosen as the simulation task. The proposed method was evaluated using a 6-DOF robotic arm holding a drill and achieved a success rate of 80% out of 20 trials.

David Ma, Kaijing Ma, Shawn Guo et al.

Recent progress in formal theorem proving has benefited from large-scale proof generation and verifier-aware training, but agentic proving is rarely integrated into prover training, appearing only at inference time. We present OProver, a unified framework for agentic formal theorem proving in Lean 4, in which failed proof attempts are iteratively revised using retrieved compiler verified proofs and Lean compiler feedback. OProver is trained through continued pretraining followed by iterative post-training: each iteration runs agentic proving, indexes newly verified proofs into OProofs and the retrieval memory, uses repair trajectories as SFT data, and uses unresolved hard cases for RL. OProofs is built from public Lean resources, large-scale proof synthesis, and agentic proving traces, containing 1.77M Lean statements, 6.86M compiler-verified proofs, and serialized trajectories with retrieved context, failed attempts, feedback, and repairs. Across five benchmarks, OProver-32B attains the best Pass@32 on MiniF2F (93.3%), ProverBench (58.2%), and PutnamBench (11.3%), and ranks second on MathOlympiad (22.8%) and ProofNet (33.2%) more top placements than any prior open-weight whole-proof prover.

Rui-Jie Zhu, Zixuan Wang, Kai Hua et al. · princeton

Modern LLMs are trained to "think" primarily via explicit text generation, such as chain-of-thought (CoT), which defers reasoning to post-training and under-leverages pre-training data. We present and open-source Ouro, named after the recursive Ouroboros, a family of pre-trained Looped Language Models (LoopLM) that instead build reasoning into the pre-training phase through (i) iterative computation in latent space, (ii) an entropy-regularized objective for learned depth allocation, and (iii) scaling to 7.7T tokens. Ouro 1.4B and 2.6B models enjoy superior performance that match the results of up to 12B SOTA LLMs across a wide range of benchmarks. Through controlled experiments, we show this advantage stems not from increased knowledge capacity, but from superior knowledge manipulation capabilities. We also show that LoopLM yields reasoning traces more aligned with final outputs than explicit CoT. We hope our results show the potential of LoopLM as a novel scaling direction in the reasoning era. Our model is available here: http://ouro-llm.github.io.

Jingzhe Ding, Shengda Long, Changxin Pu et al.

Recent advances in coding agents suggest rapid progress toward autonomous software development, yet existing benchmarks fail to rigorously evaluate the long-horizon capabilities required to build complete software systems. Most prior evaluations focus on localized code generation, scaffolded completion, or short-term repair tasks, leaving open the question of whether agents can sustain coherent reasoning, planning, and execution over the extended horizons demanded by real-world repository construction. To address this gap, we present NL2Repo Bench, a benchmark explicitly designed to evaluate the long-horizon repository generation ability of coding agents. Given only a single natural-language requirements document and an empty workspace, agents must autonomously design the architecture, manage dependencies, implement multi-module logic, and produce a fully installable Python library. Our experiments across state-of-the-art open- and closed-source models reveal that long-horizon repository generation remains largely unsolved: even the strongest agents achieve below 40% average test pass rates and rarely complete an entire repository correctly. Detailed analysis uncovers fundamental long-horizon failure modes, including premature termination, loss of global coherence, fragile cross-file dependencies, and inadequate planning over hundreds of interaction steps. NL2Repo Bench establishes a rigorous, verifiable testbed for measuring sustained agentic competence and highlights long-horizon reasoning as a central bottleneck for the next generation of autonomous coding agents.

Enduo Zhao, Murilo M. Marinho, Kanako Harada

Robotic assistance for experimental manipulation in the life sciences is expected to enable favorable outcomes, regardless of the skill of the scientist. Experimental specimens in the life sciences are subject to individual variability and hence require intricate algorithms for successful autonomous robotic control. As a use case, we are studying the cranial window creation in mice. This operation requires the removal of an 8-mm circular patch of the skull, which is approximately 300 um thick, but the shape and thickness of the mouse skull significantly varies depending on the strain of the mouse, sex, and age. In this work, we develop an autonomous robotic drilling system with no offline planning, consisting of a trajectory planner with execution-time feedback with drilling completion level recognition based on image and force information. In the experiments, we first evaluate the image-and-force-based drilling completion level recognition by comparing it with other state-of-the-art deep learning image processing methods and conduct an ablation study in eggshell drilling to evaluate the impact of each module on system performance. Finally, the system performance is further evaluated in postmortem mice, achieving a success rate of 70% (14/20 trials) with an average drilling time of 9.3 min.