Science & Mathematics

Harness development plays a crucial role in building capable AI agents. Recent approaches improve harnesses through recursive self-improvement (RSI), but their performance gains often come with higher execution costs. We introduce HarnessPro, a label-free method for evolving harnesses to improve task performance while reducing execution cost. In this setting, candidate harnesses are evaluated using LLM-generated tests without access to ground-truth solutions or official tests. Specialized agents diagnose execution trajectories, plan updates with different performance-cost priorities, and implement them as candidate harnesses. Rather than returning a single harness, HarnessPro uses normalized weighted squared-distance selection to provide cost-focused (lite), balanced (medium), and performance-focused (heavy) harnesses for different deployment budgets. Selected harnesses enter the next iteration, where execution feedback guides further updates. Our experiments cover Terminal-Bench 2, SWE-Bench Pro, and ProgramBench, using GPT-5.5 with Codex and GLM-5.2 with Claude Code. Across these benchmarks using GPT-5.5 with Codex, the heavy mode of HarnessPro achieves an average pass-rate gain of 2.0 percentage points and an average per-task cost reduction of 26.1% relative to RHO. This advantage also holds when using GLM-5.2 with Claude Code. Code is available at https://github.com/pzs19/HarnessPro

The next step for Venus exploration after the remote sensing missions of the early 2030s will be in-situ aerial platforms, the Venus equivalent of rovers for Mars. Sophisticated Venus balloon technology is already under development and presents an opportunity for the United Kingdom to act as a bilateral partner in missions exploring key questions about the volcanic and geological history of Venus, the potential habitability of the Venus clouds, and Venus as an analogue for exoplanets. White paper written for the UKSA Space Frontiers 2035 programme.

Generalized circuits are a fundamental tool for establishing PPAD-hardness and membership of approximate equilibrium problems. Despite their importance, the complexity of finding approximate solutions has remained unresolved for a wide range of approximation parameters. We settle this question by establishing a tight threshold at $\varepsilon=1/6$. Specifically, we prove that $\varepsilon$-GCircuit is PPAD-complete for every constant $01/5$, and of the fan-out-one formulations of $δ$-Pure-Circuit for every constant $δ>1/8$. These results follow from a decomposition theorem for directed graphs of maximum in-degree two and establish that the PCP-for-PPAD conjecture cannot be true for the aforementioned values of $\varepsilon$ and $δ$, unless PPAD = P. To showcase the contribution of our positive results, we apply them to a variety of problems in algorithmic game theory, including public goods games, financial networks, Bayesian, polymatrix, and graphical games, and obtain improved approximation guarantees.

For most languages, automatic speech recognition (ASR) is limited less by model architecture than by a lack of accurately transcribed spontaneous speech. We present Monsoon, a framework for collecting such data at scale. It elicits unscripted conversations between paired contributors, screens recordings with automated classifiers, corrects transcripts over five levels of human review, and selects segments for lexical and acoustic coverage. It has collected speech in 50 languages; we analyse seven 3,500-hour corpora. At equal volume, its quality stages reduce semantic word error rate (WER) by 10-39% relative to a conventional protocol. Compared with an architecture-matched baseline trained on public corpora, Whisper fine-tuned on Monsoon performs comparably on the baseline's own test sets and reduces semantic WER by 20-44 points on conversational benchmarks. The gains transfer to a second architecture and grow with data volume. We release the framework, its quality-control models, and the fine-tuned models.

Foundation models offer a promising basis for robotic agents, but long-horizon tasks require them to sustain coherent reasoning and reliable physical execution across interdependent stages. We introduce Embodied Playground, a benchmark for evaluating how foundation-model agents coordinate physical, strategic, and spatial reasoning throughout such tasks. Agents must track task progress, satisfy intermediate constraints, and revise their plans using execution feedback as earlier actions shape subsequent decisions. The playground comprises 410 task environments spanning basic manipulation, sequential puzzles, and multi-stage household activities, with variations in planning depth, geometric constraints, and physical execution demands. We curate EP-100, a representative selection of 100 tasks across five suites and 16 task types, together with the EP-Easy and EP-Hard difficulty splits and the compact EP-20 selection. A shared protocol specifies observation and tool access, resource budgets, repeated trials, and independent checks of goal completion, task-rule compliance, and physical validity. Embodied Playground provides a unified framework for studying whether foundation-model agents can sustain effective reasoning and action over extended task horizons. https://winniechen2002.github.io/Agent-bench-website/

Multi-agent systems can differ under identical agents and communication budgets when they support different forms of joint interaction. We introduce Topological Sheaves of Irreducibility (TSI), a sheaf-theoretic framework separating task-level compatibility obstructions from system capability. A task is represented by an interaction complex $K$ and a capacity sheaf $\mathcal{F}$; nonzero $H^n(K;\mathcal{F})$ defines a structural signature, while registered interventions test capability. For cocycle data, cochain realization is solvable exactly when the represented cohomology class vanishes. Examples show that the signature is sheaf-relative, and the degree-1 reduction yields cycle-space witnesses and observation requirements. We evaluate two finite one-round models. Across 15,625 degree-1 inputs, an all-agent additive primitive solves decision and repair exactly; the optimal pairwise system reaches 88.80% on decision, one input above the 88.79% majority-class baseline, and 26.63% on repair. On a degree-2 tetrahedral-boundary task over 531,441 inputs, the same-2-skeleton system retains all four three-agent primitives yet matches the pairwise-only optimum throughout the registered separation window, whereas adding the class-filling four-agent primitive makes both tasks exact. Status-preserving robustness checks retain exact performance but are not cost-matched to the class-restoring ablation. Conditional on the registered primitives and cost models, TSI provides task-relative, falsifiable tests of higher-order interaction.

General-purpose multimodal models, including GPT-6 Astra and Claude Opus 5.5, demonstrate broad capabilities in embodied manipulation. Our analysis indicates that this generality is accompanied by strong spatial perception, as Astra can determine where to interact with task-relevant entities amid visual distractors. Because spatial perception connects high-level reasoning to the physical scene, we investigate scaling visual grounding to strengthen the perceptual foundation of general embodied operation. We develop a unified data pipeline that combines robot manipulation recordings with complementary visual data from diverse sources. After automatically annotating the robot recordings, the pipeline converts the collected data into verifiable grounding supervision. Fine-tuning Qwen3.5-9B on approximately 1.5 million samples produced by this pipeline yields ACE-Brain-0.6. The model outperforms GPT-6 Astra on four of seven public grounding benchmarks, with an overall average score of 75.41 compared with Astra's 73.68. Under a fixed Qwen3.5-9B fine-tuning setup, aggregate grounding performance improves steadily as successive data releases expand the training collection to cover a broader range of grounding scenarios. These results show that scaling verifiable grounding supervision from diverse sources can strengthen the spatial perception of an open 9B model for general embodied operation.

When given an image and a text prompt, vision-language models (VLMs) can produce seemingly sophisticated text responses. However, they are also prone to hallucinate objects in their output, which can have detrimental consequences in safety-critical applications. To detect these hallucinations, certain methods interestingly leverage internal signals from the VLM as it generates the text response; potentially, these methods use the same signals that led to a hallucination to detect it. We conduct an in-depth analysis to understand the robustness of these detection methods. First, we show that a naive baseline of a generated word's token position in the VLM text response achieves competitive performance. Further, we reveal how some detection methods rely on token position and use this heuristic as a shortcut. Second, we explore detection methods' sensitivity to language and vision modalities. Our experiments suggest that strong detection performance is often driven by detecting language-based hallucinations, and when methods rely on token position, they tend to be less visually grounded and exhibit reduced robustness when hallucinations occur earlier in a text response. Lastly, we demonstrate that different detection methods can carry complementary signals; when ensembling methods together, adding weaker methods to an ensemble of higher performing ones still offers performance improvements. Altogether, while there may be useful internal VLM signals to detect hallucinations, our investigation reveals important limitations of the current methods that try to leverage them.

Egocentric video supports diverse applications, including robot learning, augmented reality, and human activity understanding. Recovering metric 3D hand trajectories from such video requires estimating both hand motion and camera egomotion to express the hands in a consistent reference frame. Many existing pipelines recover hand and camera motion in separate stages, combining hand detection, hand reconstruction, camera tracking, and trajectory refinement. We present V-JEPA Hand, a simple feed-forward video model that jointly reconstructs bilateral hand motion and camera motion from egocentric clips. It uses a self-supervised V-JEPA 2.1 encoder to extract spatio-temporal features from full-frame video and a query-based predictor to decode these features into per-frame hand poses, hand presence, and camera motion. We use the predicted camera motion to express the hand trajectories in a common world frame. The model accepts known focal lengths or predicts them when unavailable. It reconstructs each clip in a single forward pass. We train both components end to end in a single stage on a mixture of diverse pseudo-labelled videos and high-precision 3D annotations. A single model achieves the lowest coverage-aware mean per-joint position error (MPJPE) among comparable baselines on ARCTIC, HOT3D, and EgoStandard. It also yields lower world-frame trajectory errors than the compared SLAM-based pipelines. Ablations support the benefits of video pretraining, temporal context, and mixed supervision. V-JEPA Hand achieves a forward-pass throughput of 93.4 FPS on a single RTX A6000. Project page: https://qft-333.github.io/V-JEPA-Hand-Page/.

Phone scams impose substantial and growing costs on individuals and financial systems worldwide, yet existing countermeasures impose no resource cost on perpetrators. Strategic scam-baiting, in which automated agents engage scammers to deplete their operational capacity, offers a proactive alternative but requires dynamic strategy adaptation that static systems cannot provide. We introduce a game-theoretic framework enabling LLMs to learn effective counter-strategies through in-context experience without fine-tuning. The framework models scammer-baiter interactions as a two-player game with empirically grounded strategies and utility functions, using Nash equilibrium as a formal benchmark for evaluating learned behaviour. Across 300 simulated dialogues and three LLM architectures, adaptive strategy selection achieves 28.9-79.7% closer alignment to Nash equilibrium than non-adaptive baselines. The code and an annotated dataset of 300 dialogues with turn-level strategy labels and utility scores are released at https://github.com/nardinebasta/GameofPhones_/ .

Multi-chain reasoning is a challenging task not only because the multi-chain evidence is dispersed, but also because different parts of reasoning may require distinct, interacting chains of support. Flat retrieval can flood a model with loosely related text, whereas early selection of a single reasoning path can discard evidence that becomes useful only after other facts are considered. We introduce RESIN (Reasoning with Entity Structured Indexing), which uses structure to locate evidence while keeping the original source as the basis for reasoning. RESIN decomposes a question into focused one-hop needs, retrieves a small set of source-grounded units for each hop, and places only this relevant evidence in a compact reasoning buffer. The buffer preserves alternative supports until final composition, making intermediate decisions explicit without forcing a brittle path. On the full HolisQA benchmark, RESIN achieves 82.92 context recall and 76.70 answer relevancy over all questions, with higher reported CR and AR across five domains and higher F1 in four domains than the published baselines. Ablations show that both typed evidence diversity and hop-wise decomposition are essential, supporting a clear conclusion: Precise evidence organization and delayed commitment make multi-chain reasoning effective and auditable. Our code is available at https://github.com/RamonMeng/RESIN.

Traditional evaluation benchmarks reduce inherently interconnected scientific knowledge in life sciences into flat lists of questions, disregarding the underlying topological structure of the knowledge. We introduce, the first graph-structured benchmark for life sciences, featuring over 10,000 high-quality multiple-choice questions across medicine, biology, and chemistry. Our approach constructs a weighted evaluation graph using bidirectional matching and semantic similarity algorithms, where nodes represent questions and edge weights capture their semantic relationships. Leveraging this graph topology, we design two novel evaluation metrics. The Global Coherence Score (GCS) measures a model's consistency within semantically related neighborhoods, while Knowledge Balance Score (KBS) analyzes how model errors are distributed across the graph to reveal conceptual blind spots. LG-Bench facilitates fine-grained comparison of LLMs by surfacing differences in conceptual coherence and patterns of knowledge organization across models. Our framework shifts the evaluation paradigm from flat accuracy metrics to structure-aware analysis, offering a new lens for diagnosing and improving LLM performance in the life sciences domain.

A market maker provides liquidity by quoting prices to buy and sell an asset, earning the spread while managing inventory risk. We study market makers trained with multi-agent reinforcement learning on prediction markets, specifically Polymarket, whose thin order books pose distinct challenges. Three challenges underpin this research: adverse selection may cause a policy trained on real data to stop quoting; the consequences of a quote can arrive many steps after the action; and prediction-market L2 data does not identify individual orders needed for simulation. We address these challenges by reconstructing order-level LOBSTER-format data from Polymarket and training agents in JaxMARL-HFT on 417 real markets. We enrich the baseline market maker's observation space with the book and recent flow, redesign its actions and inventory controls, and propose a cooperative multi-agent decomposition into a quoter and a risk manager trained using recurrent Independent Proximal Policy Optimisation (IPPO) with RNN policies. We investigate reward shaping, particularly reward blending and post-fill markout, to mitigate credit-assignment difficulties between the agents. Our redesigned single-agent market maker quotes more frequently and incurs smaller held-out losses than the retrained JaxMARL-HFT baseline, although the contributions of jointly changed observations, actions and risk constraints are not separately identified. Decomposition alone performs poorly; careful reward design substantially improves the desk relative to the unshaped split and yields our best result. Models are evaluated on realised end-of-episode portfolio value on held-out markets. None establishes positive average portfolio value in periodic held-out evaluation, and each configuration was trained with one seed. The environment, data and protocol lay a foundation for makers that quote many linked markets at once.

Training a machine learning (ML) model is an optimization process that typically requires a set of examples with ground-truth labels beforehand. Obtaining these ground truths usually requires laborious human effort and can be costly. The scalability of human labeling becomes a concern when there are a large number of data points that need to be labeled. To address this practical issue, we consider an optimization setting where a subset of samples has ground-truth labels, while the remaining data points were labeled by an ML predictor, which could potentially assign incorrect labels to those points. We propose both first-order and zeroth-order stochastic optimization methods for training on datasets consisting of a mix of true-labeled and ML-labeled samples. We further conduct comprehensive empirical studies from training a linear regression model to fine-tuning a large language model (LLM) to showcase the effectiveness of the proposed algorithms for training with ML-labeled data.

Equations of state (EOSs) describe how a material's internal energy and pressure change with temperature and density. They are inputs to hydrodynamic simulations in materials science and planetary science. We model the ion-thermal free-energy term for dissociative materials, where molecular bonds break apart to form new species under extreme conditions. This term is governed by two latent physical processes, a molar-mass surface and a Debye-temperature curve, neither of which is observed directly. Instead, both must be recovered from internal energy and pressure, which are derivatives of the free energy, subject to thermodynamic constraints. Existing EOS tables are typically built through material-specific hand-tuning, a process that is tedious, does not scale across materials or conditions, and offers no guarantee of thermodynamic consistency. We propose a partitioned statistical framework that infers molar mass and Debye temperature directly from EOS data, guaranteeing thermodynamic consistency by construction. Local molar-mass behavior is captured with polynomial basis functions across domain partitions, while the Debye temperature is modeled with an I-spline basis. In simulations and a carbon dioxide case study, the method accurately recovers the underlying physical quantities, improves prediction of held-out energy and pressure data, and provides uncertainty quantification while scaling with EOS table size.