Abstract
Verification is critical for improving agents: it provides the reward signal for Reinforcement Learning and enables inference-time gains through Test-Time Scaling (TTS). Despite its importance, verification in software engineering (SWE) agent settings often relies on code execution, which can be difficult to scale due to environment setup overhead. Scalable alternatives such as patch classifiers and heuristic methods exist, but they are less grounded in codebase context and harder to interpret. To this end, we explore Agentic Rubrics: an expert agent interacts with the repository to create a context-grounded rubric checklist, and candidate patches are then scored against it without requiring test execution. On SWE-Bench Verified under parallel TTS evaluation, Agentic Rubrics achieve a score of 54.2% on Qwen3-Coder-30B-A3B and 40.6% on Qwen3-32B, with at least a +3.5 percentage-point gain over the strongest baseline in our comparison set. We further analyze rubric behavior, showing that rubric scores are consistent with ground-truth tests while also flagging issues that tests do not capture. Our ablations show that agentic context gathering is essential for producing codebase-specific, unambiguous criteria. Together, these results suggest that Agentic Rubrics provide an efficient, scalable, and granular verification signal for SWE agents.

Why we are recommending this paper?
Due to your Interest in Agentic RL

This paper directly addresses verification, a crucial component for reinforcement learning agents, aligning with your interest in Agentic RL. The focus on software engineering agents provides a concrete application of this verification technique.

AI Insights

• The paper discusses the concept of agentic AI, which refers to AI systems that can make decisions and take actions on their own without human intervention. [3]
• Autonomous agent: An AI system that can operate independently and make decisions based on its internal state and external environment. [3]
• Agentic AI is a rapidly growing field with various applications, including scientific discovery, language translation, and task management. [2]

Abstract
The evolution of Large Language Models (LLMs) from passive text generators to autonomous, goal-driven systems represents a fundamental shift in artificial intelligence. This chapter examines the emergence of agentic AI systems that integrate planning, memory, tool use, and iterative reasoning to operate autonomously in complex environments. We trace the architectural progression from statistical models to transformer-based systems, identifying capabilities that enable agentic behavior: long-range reasoning, contextual awareness, and adaptive decision-making. The chapter provides three contributions: (1) a synthesis of how LLM capabilities extend toward agency through reasoning-action-reflection loops; (2) an integrative framework describing core components perception, memory, planning, and tool execution that bridge LLMs with autonomous behavior; (3) a critical assessment of applications and persistent challenges in safety, alignment, reliability, and sustainability. Unlike existing surveys, we focus on the architectural transition from language understanding to autonomous action, emphasizing the technical gaps that must be resolved before deployment. We identify critical research priorities, including verifiable planning, scalable multi-agent coordination, persistent memory architectures, and governance frameworks. Responsible advancement requires simultaneous progress in technical robustness, interpretability, and ethical safeguards to realize potential while mitigating risks of misalignment and unintended consequences.

Why we are recommending this paper?
Due to your Interest in Agentic RL

This paper explores the broader evolution of agentic AI, directly relevant to your interest in reinforcement learning and autonomous systems. It examines the core elements of agentic AI, offering valuable insights into the field's trajectory.

Abstract
This work introduces an online Bayesian game-theoretic method for behavior identification in multi-agent dynamical systems. By casting Hamilton-Jacobi-Bellman optimality conditions as linear-in-parameter residuals, the method enables fast sequential Bayesian updates, uncertainty-aware inference, and robust prediction from limited, noisy data-without history stacks. The approach accommodates nonlinear dynamics and nonquadratic value functions through basis expansions, providing flexible models. Experiments, including linear-quadratic and nonlinear shared-control scenarios, demonstrate accurate prediction with quantified uncertainty, highlighting the method's relevance for adaptive interaction and real-time decision making.

Why we are recommending this paper?
Due to your Interest in Reinforcement Learning

This paper presents an online Bayesian method for behavior identification, a technique highly relevant to reinforcement learning in multi-agent systems. The use of uncertainty-aware inference aligns with the need for robust learning in complex environments.

Abstract
Reinforcement learning with verifiable rewards (RLVR) is a simple but powerful paradigm for training LLMs: sample a completion, verify it, and update. In practice, however, the verifier is almost never clean--unit tests probe only limited corner cases; human and synthetic labels are imperfect; and LLM judges (e.g., RLAIF) are noisy and can be exploited--and this problem worsens on harder domains (especially coding) where tests are sparse and increasingly model-generated. We ask a pragmatic question: does the verification noise merely slow down the learning (rate), or can it flip the outcome (fate)? To address this, we develop an analytically tractable multi-armed bandit view of RLVR dynamics, instantiated with GRPO and validated in controlled experiments. Modeling false positives and false negatives and grouping completions into recurring reasoning modes yields a replicator-style (natural-selection) flow on the probability simplex. The dynamics decouples into within-correct-mode competition and a one-dimensional evolution for the mass on incorrect modes, whose drift is determined solely by Youden's index J=TPR-FPR. This yields a sharp phase transition: when J>0, the incorrect mass is driven toward extinction (learning); when J=0, the process is neutral; and when J<0, incorrect modes amplify until they dominate (anti-learning and collapse). In the learning regime J>0, noise primarily rescales convergence time ("rate, not fate"). Experiments on verifiable programming tasks under synthetic noise reproduce the predicted J=0 boundary. Beyond noise, the framework offers a general lens for analyzing RLVR stability, convergence, and algorithmic interventions.

Why we are recommending this paper?
Due to your Interest in Reinforcement Learning

This paper tackles the challenge of verifiable rewards, a key area for improving reinforcement learning training. The focus on noisy rewards and verification strategies directly addresses the practical difficulties in RL.

Abstract
Reinforcement Learning (RL) is a well-established framework for sequential decision-making in complex environments. However, state-of-the-art Deep RL (DRL) algorithms typically require large training datasets and often struggle to generalize beyond small-scale training scenarios, even within standard benchmarks. We propose a neuro-symbolic DRL approach that integrates background symbolic knowledge to improve sample efficiency and generalization to more challenging, unseen tasks. Partial policies defined for simple domain instances, where high performance is easily attained, are transferred as useful priors to accelerate learning in more complex settings and avoid tuning DRL parameters from scratch. To do so, partial policies are represented as logical rules, and online reasoning is performed to guide the training process through two mechanisms: (i) biasing the action distribution during exploration, and (ii) rescaling Q-values during exploitation. This neuro-symbolic integration enhances interpretability and trustworthiness while accelerating convergence, particularly in sparse-reward environments and tasks with long planning horizons. We empirically validate our methodology on challenging variants of gridworld environments, both in the fully observable and partially observable setting. We show improved performance over a state-of-the-art reward machine baseline.

Why we are recommending this paper?
Due to your Interest in Deep Learning for Reinforcement Learning

This paper investigates sample efficiency in reinforcement learning, a critical factor for scaling DRL algorithms. The exploration of neurosymbolic approaches offers potential solutions to the data requirements often associated with deep reinforcement learning.