Ratchet: A Minimal Hygiene Recipe for Self-Evolving LLM Agents

Self-evolving skill libraries, pioneered by Voyager, let frozen LLM agents accumulate reusable knowledge without weight updates, yet recent evaluation shows that LLM-authored skills deliver $+0.0$pp over no-skill baselines while human-curated ones deliver $+16.2$pp: the bottleneck is not skill authoring but lifecycle management. We introduce \textbf{Ratchet}, a single-agent loop in which a frozen LLM writes, retrieves, curates, and retires its own natural-language skills. Ratchet integrates four candidate hygiene mechanisms: outcome-driven retirement, a bounded active-cap, meta-skill authoring guidance, and pattern canonicalisation. On MBPP+ hard-100 with Claude Opus 4.7, Ratchet lifts held-out pass@1 from a $0.258 \pm 0.047$ baseline to a late-window rolling mean of $0.584$ (peak $0.658 \pm 0.042$) across 100 rounds and 3 seeds, a $+0.328 \pm 0.018$ rolling-mean gain where the no-skill control drifts at $+0.002 \pm 0.005$; the same recipe transfers to an agentic solver on SWE-bench Verified ($+0.22$ peak lift over 20 rounds). Eight ablations (A1--A8) reveal that the minimal working recipe is smaller than our design suggests: retirement and the meta-skill authoring prior are load-bearing, while explicit deduplication (canonicalisation, cover-guard) is subsumed by the meta-skill itself. A non-divergence proposition shows that bounded cap and retirement threshold together prevent expected performance from drifting below the no-skills floor.

RCGDet3D: Rethinking 4D Radar-Camera Fusion-based 3D Object Detection with Enhanced Radar Feature Encoding

4D automotive radar is indispensable for autonomous driving due to its low cost and robustness, yet its point cloud sparsity challenges 3D object detection. Existing 4D radar-camera fusion methods focus on complex fusion strategies, trading inference speed for marginal gains. This trade-off hinders real-time deployment due to heavy computation on dense feature maps. In contrast, feature extraction from sparse radar points is less time-consuming but remains under-explored. This work uncovers that simply enhancing radar feature extraction can achieve comparable or even higher performance than elaborate fusion modules, while maintaining real-time performance. Based on this finding, we propose RCGDet3D, which centers on radar feature encoding and simplifies multi-modal fusion. Its encoder inherits from the efficient Gaussian Splatting-based Point Gaussian Encoder (PGE) in RadarGaussianDet3D with two key improvements. First, the Ray-centric PGE (R-PGE) predicts Gaussian attributes in ray-aligned coordinate systems before unifying them to Bird's-Eye View (BEV) space, significantly improving geometric consistency and reducing learning difficulty by decoupling the coordinate transformation from representation learning. Second, a Semantic Injection (SI) module incorporates visual cues from images, producing more geometrically accurate and semantically enriched radar features. Experiments on View-of-Delft (VoD) and TJ4DRadSet show that RCGDet3D outperforms state-of-the-art methods in both accuracy and speed, setting a new benchmark for real-time deployment.

Library Drift: Diagnosing and Fixing a Silent Failure Mode in Self-Evolving LLM Skill Libraries

Self-evolving skill libraries face a silent failure mode we term \emph{library drift}: unbounded skill accumulation without outcome-driven lifecycle management causes retrieval degradation, false-positive injections, and performance stagnation. Recent evaluation confirms the symptom--LLM-authored skills deliver +0.0pp gain while human-curated ones deliver +16.2pp (SkillsBench)--yet the underlying mechanism has not been isolated. We provide (1) a reproducible trigger: ablations that isolate drift--one disables skill injection (flat floor, +0.002), one imposes premature retirement (active harm, $-$0.019); (2) trace-level diagnostics: an append-only evidence log with per-skill contribution scores, attribution verdicts, and router engagement metrics that make the failure visible before it reaches end-task scores; and (3) a verified fix: a minimal governance recipe (outcome-driven retirement + bounded active-cap + meta-skill authoring prior) that lifts held-out pass@1 from a 0.258 baseline to a late-window mean of 0.584 (rolling gain $+$0.328) on MBPP+ hard-100 over 100 rounds. Eight ablations decompose which governance mechanisms are load-bearing and which are subsumed, providing a concrete playbook for diagnosing library drift in any self-evolving agent.

Hindsight Preference Optimization for Financial Time Series Advisory

Time series models predict numbers; decision-makers need advisory -- directional signals with reasoning, actionable suggestions, and risk management. Training language models for such predictive advisory faces a fundamental challenge: quality depends on outcomes unknown at prediction time. We bridge two ideas from reinforcement learning -- using information unavailable during execution to retrospectively generate training signal, and preference alignment -- and propose Hindsight Preference Optimization: observed outcomes let an LLM judge rank candidate advisories on dimensions that scalar metrics cannot capture, producing preference pairs for DPO without human annotation. We apply this to Vision-Language-Model-based predictive advisories on S&P 500 equity time series, demonstrated by a 4B model outperforming its 235B teacher on both accuracy and advisory quality.

Prompt Optimization Is a Coin Flip: Diagnosing When It Helps in Compound AI Systems

Prompt optimization in compound AI systems is statistically indistinguishable from a coin flip: across 72 optimization runs on Claude Haiku (6 methods $\times$ 4 tasks $\times$ 3 repeats), 49% score below zero-shot; on Amazon Nova Lite, the failure rate is even higher. Yet on one task, all six methods improve over zero-shot by up to $+6.8$ points. What distinguishes success from failure? We investigate with 18,000 grid evaluations and 144 optimization runs, testing two assumptions behind end-to-end optimization tools like TextGrad and DSPy: (A) individual prompts are worth optimizing, and (B) agent prompts interact, requiring joint optimization. Interaction effects are never significant ($p > 0.52$, all $F < 1.0$), and optimization helps only when the task has exploitable output structure -- a format the model can produce but does not default to. We provide a two-stage diagnostic: an \$80 ANOVA pre-test for agent coupling, and a 10-minute headroom test that predicts whether optimization is worthwhile -- turning a coin flip into an informed decision.

Do Agent Rules Shape or Distort? Guardrails Beat Guidance in Coding Agents

Developers increasingly guide AI coding agents through natural language instruction files (e.g., CLAUDE.md, .cursorrules), yet no controlled study has measured whether these rules actually improve agent performance or which properties make a rule beneficial. We scrape 679 such files (25,532 rules) from GitHub and conduct the first large-scale empirical evaluation, running over 5,000 agent runs with a state-of-the-art coding agent on SWE-bench Verified. Rules improve performance by 7--14 percentage points, but random rules help as much as expert-curated ones -- suggesting rules work through context priming rather than specific instruction. Negative constraints ("do not refactor unrelated code") are the only individually beneficial rule type, while positive directives ("follow code style") actively hurt -- a pattern we analyze through the lens of potential-based reward shaping (PBRS). Moreover, individual rules are mostly harmful in isolation yet collectively helpful, with no degradation up to 50 rules. These findings expose a hidden reliability risk -- well-intentioned rules routinely degrade agent performance -- and provide a clear principle for safe agent configuration: constrain what agents must not do, rather than prescribing what they should.

From Debate to Decision: Conformal Social Choice for Safe Multi-Agent Deliberation

Multi-agent debate improves LLM reasoning, yet agreement among agents is not evidence of correctness. When agents converge on a wrong answer through social reinforcement, consensus-based stopping commits that error to an automated action with no recourse. We introduce Conformal Social Choice, a post-hoc decision layer that converts debate outputs into calibrated act-versus-escalate decisions. Verbalized probability distributions from heterogeneous agents are aggregated via a linear opinion pool and calibrated with split conformal prediction, yielding prediction sets with a marginal coverage guarantee: the correct answer is included with probability ${\geq}\,1{-}α$, without assumptions on individual model calibration. A hierarchical action policy maps singleton sets to autonomous action and larger sets to human escalation. On eight MMLU-Pro domains with three agents (Claude Haiku, DeepSeek-R1, Qwen-3 32B), coverage stays within 1--2 points of the target. The key finding is not that debate becomes more accurate, but that the conformal layer makes its failures actionable: 81.9% of wrong-consensus cases are intercepted at $α{=}0.05$. Because the layer refuses to act on cases where debate is confidently wrong, the remaining conformal singletons reach 90.0--96.8% accuracy (up to 22.1pp above consensus stopping) -- a selection effect, not a reasoning improvement. This safety comes at the cost of automation, but the operating point is user-adjustable via $α$.

Multipath Extended Target Tracking with Labeled Random Finite Sets

High-resolution radar sensors are critical for autonomous systems but pose significant challenges to traditional tracking algorithms due to the generation of multiple measurements per object and the presence of multipath effects. Existing solutions often rely on the point target assumption or treat multipath measurements as clutter, whereas current extended target trackers often lack the capability to maintain trajectory continuity in complex multipath environments. To address these limitations, this paper proposes the multipath extended target generalized labeled multi-Bernoulli (MPET-GLMB) filter. A unified Bayesian framework based on labeled random finite set theory is derived to jointly model target existence, measurement partitioning, and the association between measurements, targets, and propagation paths. This formulation enables simultaneous trajectory estimation for both targets and reflectors without requiring heuristic post-processing. To enhance computational efficiency, a joint prediction and update implementation based on Gibbs sampling is developed. Furthermore, a measurement-driven adaptive birth model is introduced to initialize tracks without prior knowledge of target positions. Experimental results from simulated scenarios and real-world automotive radar data demonstrate that the proposed filter outperforms state-of-the-art methods, achieving superior state estimation accuracy and robust trajectory maintenance in dynamic multipath environments.

CC-OR-Net: A Unified Framework for LTV Prediction through Structural Decoupling

Customer Lifetime Value (LTV) prediction, a central problem in modern marketing, is characterized by a unique zero-inflated and long-tail data distribution. This distribution presents two fundamental challenges: (1) the vast majority of low-to-medium value users numerically overwhelm the small but critically important segment of high-value "whale" users, and (2) significant value heterogeneity exists even within the low-to-medium value user base. Common approaches either rely on rigid statistical assumptions or attempt to decouple ranking and regression using ordered buckets; however, they often enforce ordinality through loss-based constraints rather than inherent architectural design, failing to balance global accuracy with high-value precision. To address this gap, we propose \textbf{C}onditional \textbf{C}ascaded \textbf{O}rdinal-\textbf{R}esidual Networks \textbf{(CC-OR-Net)}, a novel unified framework that achieves a more robust decoupling through \textbf{structural decomposition}, where ranking is architecturally guaranteed. CC-OR-Net integrates three specialized components: a \textit{structural ordinal decomposition module} for robust ranking, an \textit{intra-bucket residual module} for fine-grained regression, and a \textit{targeted high-value augmentation module} for precision on top-tier users. Evaluated on real-world datasets with over 300M users, CC-OR-Net achieves a superior trade-off across all key business metrics, outperforming state-of-the-art methods in creating a holistic and commercially valuable LTV prediction solution.

VP-AutoTest: A Virtual-Physical Fusion Autonomous Driving Testing Platform

The rapid development of autonomous vehicles has led to a surge in testing demand. Traditional testing methods, such as virtual simulation, closed-course, and public road testing, face several challenges, including unrealistic vehicle states, limited testing capabilities, and high costs. These issues have prompted increasing interest in virtual-physical fusion testing. However, despite its potential, virtual-physical fusion testing still faces challenges, such as limited element types, narrow testing scope, and fixed evaluation metrics. To address these challenges, we propose the Virtual-Physical Testing Platform for Autonomous Vehicles (VP-AutoTest), which integrates over ten types of virtual and physical elements, including vehicles, pedestrians, and roadside infrastructure, to replicate the diversity of real-world traffic participants. The platform also supports both single-vehicle interaction and multi-vehicle cooperation testing, employing adversarial testing and parallel deduction to accelerate fault detection and explore algorithmic limits, while OBU and Redis communication enable seamless vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) cooperation across all levels of cooperative automation. Furthermore, VP-AutoTest incorporates a multidimensional evaluation framework and AI-driven expert systems to conduct comprehensive performance assessment and defect diagnosis. Finally, by comparing virtual-physical fusion test results with real-world experiments, the platform performs credibility self-evaluation to ensure both the fidelity and efficiency of autonomous driving testing. Please refer to the website for the full testing functionalities on the autonomous driving public service platform OnSite:https://www.onsite.com.cn.