AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time. We study strong-to-weak scaffolding: whether a stronger builder model can construct inference-time harnesses that help a weaker target model solve tasks more reliably without any parameter updates. Using four representative Theory-of-Mind benchmarks, each builder model uses 5% of the data as a validation set to iteratively refine its harness over multiple rounds, after which the finalized harness is evaluated on the full test set. Empirically, this form of test-time capability transfer is highly effective, nearly doubling average target-model performance from 0.49 to 0.91. Our analysis shows that the gains come primarily from offloading unstable model reasoning into deterministic code, benchmark-specific routing, and strict answer-format enforcement, rather than from encouraging the target model to reason more extensively or sample more broadly. We further find that builder-model reasoning effort improves harness quality monotonically, platform effects are modest relative to the builder model's own capability, and weaker target models receive the largest gains. These results suggest that inference-time harness design is an important complement to conventional training-time distillation, enabling strong models to transfer cognitive structure to weaker models without retraining.

Co-Evolution in Agentic Systems: Toward Self-Directed Evolution Beyond Human Design

Agentic systems are increasingly expected to improve after deployment, yet single-entity self-evolution is often bounded by a static learning context, such as fixed tasks and feedback. This survey focuses on co-evolution in agentic systems, a multi-component form of self-evolution in which multiple agents and their environment impose adaptive pressure on one another. To organize existing papers, we propose a progressive three-stage taxonomy that traces how the system gradually sheds human-engineered constraints. Agent--Agent Co-Evolution studies how agents adapt through dynamic peers, including adversarial, collaborative, and organizational adaptation. Agent--Environment Co-Evolution extends this loop to adaptive tasks, feedback, and interaction spaces that change with the agents. Meta Co-Evolution further explores the possibility of making the evolution mechanism itself evolvable. We also discuss open challenges in evaluating such systems, scaling them across multiple components, and keeping increasingly autonomous evolutionary processes safe and controllable. This survey provides a unified foundation for building robust and open-ended agentic systems that can improve beyond fixed human-designed paths.

Long-Horizon Embodied Decision-Making via Multimodal Memory Compression

Agents are increasingly expected to act not only as task executors, but also as decision-makers on behalf of human users. This shift requires agents to accumulate evidence over long horizons, interpret implicit user preferences, and compare multiple candidates under partial observations. In this work, we propose DunphyBench, a new benchmark for evaluating agents on long-horizon human-centered embodied decision-making, where the agent must navigate through multiple embodied housing environments and make decisions that align with multi-dimensional human preferences. Unlike standard embodied reasoning tasks that often focus on procedural planning or immediate goal completion, our setting requires agents to integrate multimodal, multi-source input into coherent knowledge that supports complex reasoning across long horizon. The evaluation results reveal that there is a substantial gap between current agents and human performance. Furthermore, our diagnosis of state-of-the-art VLM-driven agents reveals that memory management is one of the bottlenecks, where raw multimodal history introduces noise that hinders decision quality. Motivated by this finding, we design MeMento, a preference-conditioned multimodal memory compressor that selectively compresses decision-relevant information from long-horizon history based on user preferences with a fixed set of memory tokens. Experiments show that MeMento helps VLM-driven agents improve accuracy by 7.18%, while reducing memory usage by 85.38% compared to the strongest baseline.

Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement Learning

Test-time scaling through iterative self-evolution with environment feedback, as demonstrated by AlphaEvolve, shows remarkable performance gains. We hypothesize that the success of such evolution frameworks hinges on meta-skills, such as self-reflection with environment feedback, that enable effective multi-round refinement, yet are largely neglected by traditional post-training. To bridge this gap, we present MetaEvolve, a framework designed to develop these meta-skills via a data synthesis pipeline, evolution-aware reinforcement learning (RL), and inference-time evolutionary search. Concretely, we ground MetaEvolve in coding, where program execution provides natural, continuous reward signals beyond binary correctness. Building on these signals, we synthesize evolution trajectories as training data, each containing a current program, its fitness score (combining correctness and efficiency), and a history of prior attempts, and train the model via RL with verifiable rewards derived from test case execution. By training on large-scale code data, we aim to inspire generalizable domain-agnostic meta-skills that can transfer broadly to open-ended problems where such rich training signals are scarce. Across seven coding benchmarks, MetaEvolve outperforms the strongest baseline by 10.01% absolute on in-distribution tasks and 24.12% on out-of-distribution tasks. On open-ended algorithm optimization problems entirely outside the training domain, it further achieves a 46.9% relative improvement. These results demonstrate that explicitly cultivating self-evolution meta-skills offers a principled path toward more capable and autonomously self-evolving AI.

Removable Defects: The Economics and Limits of Deliberate Deficiency

A specialist tolerates blind spots that a generalist does not. Usually this is treated as a cost to be minimized. We treat it as a design variable: a deficiency can be kept because it pays and removed on demand in the rare situation where it would be fatal, by routing to a compensation channel. We give three results. First, an advantage condition under which keeping the deficiency is a computable economic position; structurally it is the Ehrlich-Becker market-vs-self-insurance margin applied to a competence gap, with the detector as a Townsend costly-state-verification technology. Second, a two-sided characterization of removability. A coupling lemma shows that when the deficiency is a coarsening of perception, no switch can separate benefit from harm, yielding a converse (a confounded detector earns zero premium, and any within-defect policy insisting on positive premium is driven, under multiplicative dynamics, to negative long-run growth) and an achievability result (a detector outside the deficiency earns a positive premium). Together, over structured uncertainty classes with severity capped or miss rate O(1/L): a defect is profitably removable iff the detector-relevant distinction survives the restriction and the advantage condition holds; the premium is the support function of the class's ROC set at an economic price vector. Third, observation defects and capacity defects differ exactly on whether access to the deployment distribution rescues them; the gap decomposes as cross-leak plus a closure deficit, and per-task randomization buys back the latter, never the former. The detector can be learned from declared fatal categories at a training bill linear in loss severity (up to a log factor). The results synthesize Chow's reject option, Kelly growth under ruin, and selective prediction.

Information Limits and Attractor Dynamics in Economies of Frontier LLM Agents: A Pre-Registered Test

We report a pre-registered, two-part experiment on small economies of frontier language-model agents (Claude Opus 4.8), testing two quantitative predictions about coupled multi-agent systems: an information-theoretic capacity region for wealth growth under market coupling, and a mean-field residual-scaling law for population misalignment under incentive and control levers. All predictions, acceptance bands, and decision rules were frozen in a public git chain before any run; every reported number re-derives mechanically from cached model outputs; the entire experiment cost $138.76 in metered API spend and is re-runnable at zero cost from the cache. Result 1 (confirmation): in parimutuel-coupled economies, relative growth equals relative claimed information -- the gap law G_a - G_b = I_a - I_b holds to a worst-case 46 millinats (pre-registered band: 50) across four perception structures; coalition value is submodular exactly where channels are conditionally independent, and a designed XOR synergy control flips it supermodular by 0.62 >= ln2/2 nats, with agents reasoning out the joint bit; the joint growth ceiling G_S <= H(X) binds exactly; and the best-informed agent absorbs essentially the whole wealth pool in 4/5 market seeds. Result 2 (structural negative): the residual-scaling test returned "domain not found." In all 72 population runs, goal dispersion collapsed (V -> 0; maximum 4.85 against a frozen floor of 5.31), the population's response to the two levers was a step function across the dominance boundary rather than a smooth response, and cells near the boundary were bistable with seed-selected outcomes. No tested LLM population at any capability level realizes the noise-maintained-dispersion regime the smooth mean-field model assumes. We release the full protocol, pre-registration chain, call cache, and analysis code.

MS-Resampler: Multi-Scope Visual Resampling for Efficient Multimodal LLMs

Multimodal large language models (MLLMs) typically employ resampling-based projectors to transform dense visual features into a compact token sequence for language modeling. Most existing resamplers adopt a single, fixed aggregation scope via global cross-attention, which can blur fine-grained local evidence and limit the ability to capture both local details and global context within a fixed token budget. In this work, we propose MS-Resampler, a multi-scope visual resampling framework for MLLMs. MS-Resampler instantiates multiple scope-specific resamplers by injecting explicit spatial scope priors into the resampling attention, enabling each branch to aggregate visual information at a particular granularity from local to global. The outputs of these scope-specific resamplers are then adaptively fused to produce the final visual representations for language modeling. Extensive experiments on ten public multimodal benchmarks show that MS-Resampler consistently improves visual understanding and multimodal reasoning over conventional single-scope resamplers, while introducing only minimal computational overhead.

Trimming the Long-Tail of Visual World Modeling Evaluation

Physical interactions follow a long-tailed distribution: a set of common and regular interactions dominates human experience and visual data, while a broad spectrum of rare and irregular interactions remains underrepresented. Although recent visual world models, including image and video generation models, achieve impressive realism on existing benchmarks, they primarily focus on simulating common physical interactions. This raises a central question: Do current visual world models internalize and generalize physical principles? In this work, we introduce Tailor-Bench, a benchmark that challenges world models to simulate irregular physical interactions. To enable systematic evaluation, we design three scenario modes that progressively challenge model reasoning: Regular scenarios reflect common tool-task pairs, Unconventional scenarios replace conventional tools with attribute-compatible substitutes to test affordance generalization, and Impossible scenarios introduce attribute-violating tools to probe constraint awareness. Additionally, we design two complementary settings under a unified evaluation protocol: predictive generation requires inferring outcomes without guidance, while descriptive generation specifies the target outcome for faithful realization. Our experimental results reveal a clear long-tail gap in physical world modeling: performance degrades from Regular to Unconventional and Impossible scenarios, indicating limited generalization beyond common interactions. Failure analysis further shows that models rely on superficial visual patterns: image models fail to realize correct state changes, while video models further suffer from temporal inconsistencies.

PlanBench-XL: Evaluating Long-Horizon Planning of LLM Tool-Use Agents in Large-Scale Tool Ecosystems

LLM agents increasingly operate in large tool ecosystems, where real-world tasks require discovering relevant tools, inferring implicit sub-goals, and adapting to dynamic environments over long horizons. However, existing benchmarks rarely evaluate planning under retrieval-limited tool visibility. To address this gap, we introduce PlanBench-XL, an interactive benchmark of 327 retail tasks over 1,665 tools that tests whether agents can iteratively retrieve usable tools, invoke them to uncover intermediate evidence for subsequent calls toward the final goal. PlanBench-XL further features an optional blocking mechanism that simulates real-world unpredictability through missing, failing, or distracting tool functions, forcing agents to detect disrupted paths and adapt at runtime. Experiments on ten leading LLMs show that massive-tool planning remains challenging: while GPT-5.4 achieves 51.90% accuracy in block-free settings, it collapses to 11.36% under the most severe blocking condition. Further analysis shows that agents are especially vulnerable when failures lack explicit error signals or when recovery requires longer alternative tool-use paths. These results establish PlanBench-XL as a testbed for diagnosing agentic planning failures and highlight the need for robust adaptive planning in long-horizon tasks with large, imperfect tool environments.

ACCORD: Action-Conditioned Contextual Grounding for Language Agents

User instructions are often underspecified because humans rely on implicit assumptions about the surrounding environment. For large language model (LLM) agents operating in information-rich digital and physical environments, these assumptions cannot be inferred from the instruction alone; they must be recovered from the current state of tools, data, interfaces, and observations. Effective execution therefore requires agents to identify missing context, ground it in observed evidence, and carry it forward into subsequent actions. We show that current agents often fail to do so. They act from assumed rather than observed specifics, overlook information they could have gathered, and fail to incorporate evidence that has already been returned. Building on this insight, we propose ACCORD (Action-Conditioned Contextual Grounding), a simple and effective agent framework for adaptive grounding. Before each action, ACCORD actively probes the environment for missing information and integrates relevant context from the agent's trajectory that would otherwise be overlooked. Requiring no additional training or task-success signals, ACCORD improves task-goal completion on AppWorld by up to +20.6 points with GPT-5-mini, from 42.0% to 62.6%, compared to strong baselines. These gains persist with a substantially stronger base model (+10.8 with Claude-4.5-sonnet), an open-weight model (+10.1 with Qwen3.5-27B-FP8), and on the embodied AlfWorld benchmark (+7.4 success rate with GPT-5-mini).