This paper formalizes agentic retrieval-generation loops as finite-horizon partially observable Markov decision processes, explicitly modeling their control policies and state transitions, and develops a comprehensive taxonomy and modular architectural decomposition that categorizes systems by their planning mechanisms...
Large reasoning models achieve strong performance through reinforcement learning but can suffer safety degradation. External teacher distillation can restore safety yet may compromise reasoning performance. We formalize safety realignment as a KL projection: the student's native filter-accepted distribution is uniquely...
Seanie Lee, Sangwoo Park, Yumin Choi et al.· 0 citations
We present Ax-Prover, a multi-agent system for automated theorem proving in Lean that can solve problems across diverse scientific domains and operate either autonomously or collaboratively with human experts. To achieve this, Ax-Prover approaches scientific problem solving through formal proof generation, a process th...
Benjamin Breen, Marco Del Tredici, Jacob McCarran et al.· 0 citations
In this paper, we propose a hybrid framework for Satisfiability Modulo the Theory of Nonlinear Real Arithmetic (SMT-NRA for short). First, we introduce a two-dimensional cell-jump move, called \emph{2D-cell-jump}, generalizing the key operation, cell-jump, of the local search method for SMT-NRA. In contrast to the exis...
Tianyi Ding, Haokun Li, Xinpeng Ni et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
A universal controller for any robot morphology would greatly improve computational and data efficiency. Steps have been made towards such multi-robot control by utilizing contextual information about the properties of individual robots and exploiting their modular structure in the architecture of deep reinforcement le...
Laurens Engwegen, Max Weltevrede, Caroline Horsch et al.· 0 citations
This work introduces ID-WM (Implicit Dynamics World Models), a hierarchical world model that decouples state-space reachability from local control and yields a 135% performance gain over single-level world model MPC, particularly in complex, unseen configurations for long-horizon tasks.
Gao-Yue Zhou, Yvonne Wu, Zichen Jeff Cui et al.· 0 citations
RoboArchitect shows that harness design can be automated, offering a strong alternative to finetuning on a novel setup, and its design can be effectively automated and optimized without weight updates.
Sheng-Guang Wu, Su-Ning Huang, Cherie Ho et al.· 0 citations
This paper derives an asymptotically subquadratic sparse attention method based on a hierarchical, tree-structured indexer, which it is demonstrated can be cheaply applied to existing pretrained models, and maintains performance on long-context retrieval and challenging reasoning, mathematical, and agentic coding evalu...
Leon Lufkin, Cutter Dawes, Alexander Speicher et al.· 0 citations
Reinforcement learning with verifiable rewards (RLVR) has become the standard method for training language models to reason. However, RLVR typically relies solely on outcome correctness, which does not guarantee that the CoT causally contributes to the final answer. Without this connection, outcome rewards may fail to...
We study the stochastic second-order oracle complexity of solving smooth convex finite-sum optimization with $n$ components. Very recently, Tsedenov \& Kovalev (2026) showed that the minimax optimal complexity under averaged smoothness and convexity is $\tilde{\Theta}(n + n^{6/7} \epsilon^{-2/7})$ by establishing nearl...
Xinliang Zhang, Lesi Chen, Zonghan Li et al.· 0 citations
We study how to learn a generative model of clean images using only noisy observations. Diffusion processes and their generalizations to non-Gaussian noise require posterior predictions at every noise level. We first show that it is possible to learn these predictions from independent noisy views using Noise2Noise regr...
Fabian Kruse, Lorenzo Baldassari, Ivan Dokmani\'c· 0 citations
Looping re-applies a block of layers so that each token passes through the block more than once. Adding a loop to a trained model is attractive because it leaves the parameter count, and hence the disk size, the memory footprint and the training cost unchanged. Unfortunately, every generated token pays for the loop. Re...
