arXiv subjects

Mengsha Huang

Publications and source records attributed to Mengsha Huang.

3 recordsLinked to original sources

arXiv2026

Praxist: From Experimental Artifacts to Solution Lineages

Autonomous R\&D agents now write, run, and improve executable artifacts under automated evaluation---but largely as laboratory instruments: shown on curated benchmarks, with gains that are hard to trace to a cause and costs well above what sustained engineering practice absorbs. The limitation is structural. Most systems treat each attempt as nearly self-contained, so logs, memories, and search trees record what happened without establishing which design element produced an improvement, whether its evidence survived validation, or how it recombines with others. Long campaigns therefore keep re-learning the same lessons. We introduce Praxist, a lineage-centered generational system that converts reproducible artifacts and evaluator outcomes into a typed evidence graph of findings, lane-structured frontiers, and agendas. Separating local artifact construction from cohort-level evidence synthesis lets later attempts inherit validated mechanisms, unresolved claims, and useful constraints, and leaves results attached to an inspectable lineage. On the standardized 75-task MLE-bench suite, the finalized official-grader results give Praxist 60 medals (80.0\%), 49 of them gold, against 55 medals (73.3\%) and 34 gold for a Claude Code baseline on Claude Opus 4.8---at a recorded model spend of US\$3,054 versus US\$38,370, roughly a twelfth of the cost. Four case studies---quantitative trading, LiDAR-inertial-visual SLAM, tokamak magnetic control, and rocket landing---carry the same process into open-ended engineering problems, improving on each task-native baseline in headline accuracy, survival, or resource cost, with the discovery path on record. Stronger artifacts at an order of magnitude less spend, each backed by an auditable lineage, are, to our knowledge, first brought together here: the operating profile production research requires, not the one a benchmark demonstration establishes.

cs.MA↗

arXiv2026

Dynamics in a Low-Rank Separable Field Cellular Automaton

Cellular automata (CA), computational models used to study the emergence of life-like behaviors, classically apply local interaction and updating rules. Yet, another interacting paradigm observed in life, the globally coupled computation is usually overlooked in traditional CA. Here, we introduce a Separable-Field Cellular Automaton (SFCA), a normalized-field cellular automaton in which local-neighborhood interactions is replaced by a global coupling field. Each cell is updated according to a normalized field, with survival and birth governed by two threshold intervals. Systematic scans over interval widths and positions revealed four outcome classes: extinction, fixed points, cycles, and long transients. The outcome phase diagram was organized by the relative geometry of the survival and birth intervals: fixed points dominated when born interval was contained in survival interval, whereas long transients concentrated near the boundary between partial overlap and no overlap. A fine scan along this transition showed that the long-transient region forms a narrow but persistent ridge separating two qualitatively distinct cycle-dominated regimes. One side produced dense, high-change-rate cycles approximating global period-2 alternation, whereas the other produced sparse, low-change-rate, stripe-like cycles. Damage-spreading further supported a basin-competition interpretation, in which the long-transient ridge reflects delayed selection between two cyclic attractor families rather than random nonconvergence, while finite-size analysis shows that the long-transient ridge remains robust across tested grid sizes. These results illustrate that long-transient dynamics can arise under separable field coupling with a rank-1 field matrix, suggesting that nontrivial collective organization does not necessarily require complicated interaction rules in a globally coupled system.

nlin.CG↗

arXiv2026

Unraveling the Hidden Dynamical Structure in Recurrent Neural Policies

Recurrent neural policies are widely used in partially observable control and meta-RL tasks. Their abilities to maintain internal memory and adapt quickly to unseen scenarios have offered them unparalleled performance when compared to non-recurrent counterparts. However, until today, the underlying mechanisms for their superior generalization and robustness performance remain poorly understood. In this study, by analyzing the hidden state domain of recurrent policies learned over a diverse set of training methods, model architectures, and tasks, we find that stable cyclic structures consistently emerge during interaction with the environment. Such cyclic structures share a remarkable similarity with \textit{limit cycles} in dynamical system analysis, if we consider the policy and the environment as a joint hybrid dynamical system. Moreover, we uncover that the geometry of such limit cycles also has a structured correspondence with the policies' behaviors. These findings offer new perspectives to explain many nice properties of recurrent policies: the emergence of limit cycles stabilizes both the policies' internal memory and the task-relevant environmental states, while suppressing nuisance variability arising from environmental uncertainty; the geometry of limit cycles also encodes relational structures of behaviors, facilitating easier skill adaptation when facing non-stationary environments.

cs.LG↗