This paper studies policy optimization for long-horizon software-engineering agents that learn from sparse binary test outcomes. The core idea is a training recipe—dubbed SPO++—that combines the Muon optimizer with single-rollout asynchronous updates, truncated importance sampling (TIS), token-averaged loss, and auxiliary length/tool-error penalties. The authors claim this combination achieves higher training success than SPO baselines on repository-level issue resolution for two model families, suggesting Muon is critical for long-horizon RL.

Doc2Harness evaluates whether frontier LLMs can engineer a coding-agent harness from scratch, a capability that existing SWE benchmarks ignore by holding the harness fixed. The benchmark asks a build model to generate a complete harness from a development document and an interface-only skeleton, then evaluates it deterministically with a scripted mock model across $328$ black-box test cases. The results reveal a $26$-point performance gap between the best and worst models, with even the strongest scoring only $53.05\%$, suggesting substantial room for improvement in long-horizon infrastructure construction.

Document understanding has been evaluated through isolated silos—layout detection, OCR, reading order, and image captioning—making it impossible to audit whether a single coherent page parse exists. This paper proposes Read-Parse-Describe, a unified benchmark built from 2,402 ACM scientific documents that forces models to jointly output bounding boxes, semantic labels, reading order, recognized text, caption links, and alternative-text descriptions for visual elements. To combat spatial drift in compressed vision-language tokens, the authors introduce SARA, a lightweight residual adapter that re-injects crop-aligned multi-layer visual features into frozen VLMs.

Current social agent benchmarks reward fluent conversation but ignore whether agents actually reach valid plans that respect hidden constraints. This paper introduces HICO-Bench, a benchmark where two agents with private information must coordinate verifiable outcomes, and PaCPO, a reinforcement learning method that redistributes terminal rewards across turns by checking tentative plans against paired worlds. The work matters because it reveals that high LLM-judged conversation ratings frequently mask infeasible or unsupported agreements, and provides a concrete credit-assignment mechanism to train agents on verified social outcomes rather than just dialogue style.

Tabular foundation models learn to predict from in-context examples without task-specific fine-tuning, but dense architectures apply the same computation to every feature regardless of statistical context. OrionMoE-Small attacks this limitation by inserting a dataset-conditioned mixture-of-experts layer along the feature axis of a dual-axis Transformer, allowing the model to route different features to different expert networks based on both local token content and a global dataset register. If the sparse design truly helps, it could offer a principled way to scale tabular models by allocating computation heterogeneously; however, the paper never proves the MoE itself is responsible for the gains.

The paper studies how much context a Transformer must retain to perform in-context temporal-difference (TD) policy evaluation without parameter updates. Existing theoretical constructions store the full trajectory, costing Θ(Nd) memory. The authors prove the intrinsic limit is the Krylov dimension k_L, construct a fixed-parameter attention network that compresses raw transitions to Θ(nk_L) registers, and give tight worst-case approximation error Θ(γ^m max{1,1/(m(1−γ))}) when the context falls below k_L.

LLM entity extraction from full-length documents is widely deployed but its failures are poorly diagnosed. This paper runs a preregistered audit of two low-cost LLMs and a flagship control on 315 real-world cyber-threat-intelligence reports, finding that most false positives are background mentions rather than hallucinations, and that confidence signals can rank outputs for human review but cannot verify truth against relation-level counterfactuals. The results set a concrete deployment boundary for any high-stakes pipeline that triages extraction by confidence.

MUCO proposes manipulation units (MUs) to decompose long-horizon robot manipulation into reusable, verifiable sub-operations. Each MU carries a local goal, an initiation condition, and observable success/failure criteria; the MUCO framework matches MUs to specialist policy executors from a frozen capability registry, enables local retry and fallback, and avoids replaying already-verified prefixes. The conceptual goal is to reduce data-collection costs, enable cross-task reuse, and contain failures without joint training or shared backbones. The critical caveat is that the submitted preprint contains almost no quantitative results.

This paper studies how to convert U.S. state legislative PDFs into exact, schema-constrained XML. The authors test the hypothesis that exact reproduction relies more on harness engineering (tools, orchestration, deterministic constraints around a model) than on model scale. They cross three Claude models with three harness tiers (zero-shot, single-agent with tools, and decomposed multi-agent pipeline) on 153 gold bills. The headline finding is striking: the weakest model in the full pipeline outperforms the strongest model zero-shot ($0.948$ vs $0.806$), and the model spread collapses from $0.68$–$0.81$ to $0.95$ under full scaffolding.

Explorer-Zero tackles the data-curation bottleneck for agentic RL by training a learned proposer to explore stateful environments, compose multi-step tasks from demonstrated transitions, and record intermediate environment states as verifiable checkpoints. These checkpoints supply fine-grained rewards $R^{\text{subgoal}}$ that prevent group-normalized advantages from collapsing when the solver fails, while alternating proposer–solver training adapts the curriculum automatically. On ALFWorld, ScienceWorld, and AppWorld, the approach outperforms strong self-play baselines and narrows the gap to human-data RL.

Medical image parsing requires a single model to produce heterogeneous outputs—categorical labels, bounding boxes, counts, numeric estimates, and free-text reports—across disparate imaging modalities. This work studies task-specific adaptation of a shared 4-bit Qwen3.5-9B backbone using lightweight LoRA adapters, where a question-only GLiClass router selects the adapter, image budget, and prompt for each case on the FLARE 2026 Task 3 benchmark. The core finding is that adaptation recipes should vary by task—rank, resolution, and prompts allocated according to demand—rather than applied uniformly.

This paper studies the implicit bias of the Muon optimizer when its idealized spectral-norm steepest-descent update is replaced by the standard practical implementation: finite Newton–Schulz iterations with a positive stabilization constant. The authors prove that, contrary to the spectral max-margin behavior established for idealized Muon, the practical variant instead converges in direction to the Frobenius max-margin separator—effectively recovering the implicit bias of ordinary gradient descent. They further characterize how the iteration count, stabilization, and polynomial choice interact, and provide conditions (logarithmically growing iterations) under which spectral bias can be recovered.

The paper introduces Sparse Transformer (ST), a 5-layer language model trained on TinyStories that enforces exact activation sparsity in its MLPs via a TopK gate. By allowing only K neurons to fire per token, the model creates a known reference basis that serves as ground truth for evaluating mechanistic interpretability methods. The authors use this testbed to compare feature-recovery techniques (SAEs vs. Transcoders) and to study how compression induces polysemanticity, releasing all 35 trained models and checkpoints to support future research.

Sequence models usually require costly backpropagation and large datasets, leaving open the question of how far simple random reservoirs can go in few-shot settings. This paper systematically tests Echo State Network variants on the STREAM benchmark of synthetic seq2seq tasks, showing that orthonormal input initialization preserves a quasi-linear memory regime and that a proposed data-dependent gate (G-ESN) together with shared deep readouts pushes the solvable frontier toward structural tasks like adding and bracket matching. By benchmarking these cheap reservoir models against gradient-based SOTA, the study offers a practical map of when random features suffice and when fully trained models remain necessary.

Agent ecosystems rely on reusable skills that compose into complex workflows. CoRedSkill decomposes a legitimate workflow into interdependent skills and injects distributed attack operations via AST rewriting, preserving original utility while connecting malicious steps through intermediate artifacts. On a 150-case benchmark spanning exfiltration, tampering, and privilege violation, it achieves 98.0% attack success and 90.7% joint success against both skill-package and runtime auditors. The work exposes a severe gap: skill compositions can simultaneously complete legitimate tasks and carry out stealthy unauthorized operations.

Long-horizon multi-agent mathematical research suffers from fragmented context, unsupported claim propagation, and opaque reasoning trails that are difficult for humans to inspect. Euler tackles this by representing the research process as persistent, shared research graphs—recording obligations, candidate routes, blockers, and evidence—that both agents and researchers manipulate through a unified workbench. The goal is to make extended mathematical investigations traceable, resumable, and open to focused human intervention at any step.

GUI agents are increasingly deployed to automate mobile tasks, but most benchmarks only measure whether a task finishes, not whether the execution path was stable. GraceBench fills this gap by representing each app's reachable interaction space as a UI Transition Graph (UTG) and aligning video-recorded trajectories to it, enabling step-level diagnosis of deviation, looping, and recovery. The authors evaluate eleven models on 169 tasks across ten Android apps and show that success metrics hide serious instability: over half of successful trajectories still contain a deviation or loop, and even the highest-success model recovers poorly. The work matters because it provides a concrete framework for measuring process-level reliability, which is essential for testing and real-world deployment.

Typed decision models route requests or select tools from predefined categorical options, using fallbacks such as "Other" or "Defer" to handle out-of-scope inputs. This paper introduces the Paired Fallback Robustness (PFR) Test to check whether models correctly switch to fallback when required and whether they remain consistent under logically equivalent rule descriptions such as Union versus Complement definitions. The authors find that fallback mechanisms frequently fail silently: Intern's accuracy on TREC50 collapses from 97.9% to 23.1% on correctly classified inputs when reassigned to fallback, and equivalent definitions can flip individual decisions without changing aggregate accuracy. These results matter because fallback options, intended as reliability safeguards in production routing and delegation systems, can themselves become a source of error.

Current agentic systems are built by single developers with full control, which limits scalability. This paper proposes multi-developer agentic systems (MDAs) as a new scaling axis, where independent developers contribute specialized agents that are composed via task decomposition and routing. On the diverse GAIA benchmark, the authors show that a coalition of specialists—none exceeding 35% accuracy alone—reaches 70% when combined in an MDA, while a narrow-domain SealQA experiment reveals overhead costs can erase gains. A Tullock-contest theory of developers further argues that MDAs incentivize specialization, whereas single-developer routing pushes generalism.

Long-horizon agent benchmarks like SWE-bench Verified can cost thousands of dollars and days of compute per evaluation run. This paper asks whether LLM agents can automatically construct cheaper proxy benchmarks that preserve the rankings and score distributions of expensive reference evaluations. The authors propose EvalBuilder, a meta-evaluation framework that assesses generated benchmarks along fidelity, efficiency, and robustness, with explicit tests for generalization to held-out models and freshly collected responses. The goal is to reduce the human effort and computational bottleneck of manual benchmark design.