Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (CS.AI) 2026-06-19

AI4SE and SE4AI Exploration: A Decade Looking Back and Forward

arXiv:2606.19630v1 Announce Type: new Abstract: The March 2020 INCOSE INSIGHT special issue on AI and Systems Engineering (SE) became the most downloaded issue in the publication's history and launched a research community that now draws over 250 registrants to its annual workshop. In this article, we trace the progress in AI and SE across three phases (labeled here foundational, applied, and LLM inflection) based on the authors' reading of the field's core papers, and describe our opinions of where the community has converged and where critical gaps remain. Separately, a human-AI agreement literature review leveraging both human expertise and six AI models was performed to assess the relevance of 1,712 INCOSE INSIGHT articles and 889 SERC publications. The results identify five critical research gaps and offer guidance for practitioners navigating AI adoption, assurance, and workforce transformation in SE. We share the agreement data and the AI4SE/SE4AI Explorer web application so readers can compare their own relevance judgments with the human and AI raters.

02.
arXiv (CS.AI) 2026-06-16

The Energy Blind Spot: NVIDIA's Flagship Edge AI Hardware Cannot Support Process-Level Energy Attribution

arXiv:2605.27599v2 Announce Type: replace-cross Abstract: Agentic AI workloads - where a single user goal triggers multi-step orchestration, tool calls, retries, and failure recovery - are being targeted for edge deployment, with NVIDIA, Dell, HP, ASUS, MSI, Acer, and Gigabyte all shipping GB10-based desktop AI systems in 2026. We recently demonstrated that orchestration structure dominates agentic energy cost, with workflows consuming 4.33x more energy per successful goal than linear baselines and OOI reaching 7.63x for multi-step reasoning tasks. Separately, Raj et al. show that CPU-side processing accounts for up to 90.6% of total latency and 44% of total dynamic energy in agentic workloads. We report a systematic energy-observability audit of the ASUS Ascent GX10 (GB10 SoC) and find that the platform exposes no CPU energy counter, no INA power-rail monitor, no IPMI/BMC, and no SCMI powercap protocol through any supported software interface. The only on-device energy telemetry is instantaneous GPU power via NVML. We further discover that the MediaTek firmware already computes per-rail energy internally via an undocumented ACPI interface (SPBM), but NVIDIA states there are "no plans to expose CPU rail information." On-device per-process energy attribution - as performed on x86 via RAPL - is therefore not reproducible on this platform through supported interfaces. We formalize a hardware requirements specification for energy-attributed AI, propose an interim calibration bridge for per-domain energy decomposition - confirmed on the Acer Veriton GN100 where CPU energy accumulators are live - and identify a standards-track path via SCMI powercap. Our findings motivate the low-carbon computing community to demand energy observability as a first-class hardware requirement.

03.
arXiv (quant-ph) 2026-06-15

Strategic Non-Shareability of Quantum Correlations

Authors:

arXiv:2605.25516v2 Announce Type: replace Abstract: Correlations distributed by a mediator can be useful for coordination but vulnerable to inheritance by a colluder. We formalize the obstruction to such inheritance as a source-certified resource theory of strategic non-shareability. The free objects are symmetrically extendible sources, the free operations are shareability-preserving maps, and the trace distance to the free set is a faithful convex monotone. For Werner and isotropic sources in arbitrary local dimension, the resource has the exact form $D_m=c(d)(p-p_m^{*})_{+}$, with $p_m^{*}$ the Johnson–Viola shareability threshold. For qubit Werner sources, tomographically complete Pauli measurements yield the exact one-colluder capacity\[ C^tomo_1(p)=\frac{1}{12}\Bigl[(3p-1)-\sqrt{(3p+1)(1-p)}\,\Bigr]_{+}.\] We prove that this anti-collusion resource is independent of Bellnonlocality: the Bell and shareability orderings cross, so some Bell-nonlocal states are strictly less collusion-resistant than Bell-local ones. Finally, we give an aligned Pauli coordination game whose observed behaviour has a local hidden-variable model for every visibility, making device-independent certification empty, while source-certified quantum anti-collusion is positive exactly above the extendibility threshold. These results identify symmetric non-extendibility, rather than Bell nonlocality, as the boundary of source-certified collusion resistance.

04.
arXiv (CS.AI) 2026-06-16

From Detection to Recovery: Operational Analysis on LLM Pre-training with 504 GPUs

arXiv:2605.09370v5 Announce Type: replace-cross Abstract: Large-scale AI training is fundamentally a distributed systems problem, where hardware failures are routine operating conditions rather than rare exceptions, yet public operational evidence from production training clusters remains limited. This report presents an empirical analysis of a 63-node NVIDIA B200 production cluster (504 GPUs), using 55 days of Prometheus time-series data and 73 days of operational logs covering 224 multi-node training sessions. The environment is cross-organizational: five parties (SKT, Upstage, Lablup, NVIDIA Korea, VAST Data) share a unified monitoring pipeline. This enabled joint diagnosis of a 60-node-scale storage I/O bottleneck absent in 2-4-node tests, a production-scale phenomenon no single team could isolate alone. We perform three quantitative analyses yielding four findings. First, over 751 Prometheus metrics and 10 XID-identified GPU failures, no single metric is consistently dominant across failure types, motivating multi-signal detection. Second, 523 checkpoint events trace the save/load path from GPU VRAM to the NFS server: restart loading reaches 21.5% of maximum read bandwidth (700 GB/s) and save bursts 16.0% of maximum write bandwidth (250 GB/s), with NFS/RPC queueing and transport-layer backlog rising together. Third, across 224 sessions over 73 days, node exclusions concentrate so the top 3 of 63 nodes account for over 50%. Fourth, auto-retry chain analysis shows a 33.3% success rate over 12 chains (73 attempts), 2.7x the 12.5% manual rate, with a median retry interval of 11 minutes (IQR 10-11). All analyses are grounded in production infrastructure providing session-level workload management, GPU-centric scheduling, and unified observability.

05.
arXiv (CS.CV) 2026-06-12

Measurement Plasticity: Sensor-Level Adaptation for Vision-Language Models

We propose Multi-View Physical-prompt (MVP) for Test-Time Adaptation (TTA), a forward-only framework that moves TTA from tokens to photons by treating the camera exposure triangle (i.e., ISO, shutter speed, and aperture) as physical prompts. At inference, MVP acquires selected multiple physical views using a source-affinity score, evaluates digitally augmented variants of each retained view and filters the lowest-entropy predictions, and aggregates predictions with hard voting. This selection-then-vote design is simple, calibration-friendly, and requires no gradients or model modifications. On ImageNet-ES and ImageNet-ES-Diverse, MVP outperforms digital-only TTA on both Auto-Exposure and a combination with conventional sensor control. MVP remains effective under reduced parameter candidates that lower capture latency, demonstrating its practicality.

06.
arXiv (CS.CL) 2026-06-16

SAMark: A Self-Anchored Text Watermarking with Paragraph-Level Paraphrase Robustness

Semantic-level watermarking (SWM) improves robustness against text modifications by treating sentences as the basic unit. However, robustness to paragraph-level paraphrasing remains difficult because such attacks globally disrupt watermark signals by changing sentence order. In this work, we propose SAMark, a self-anchored watermarking framework that removes the dependency on sentence order by establishing a step-independent green region in semantic space. To improve detectability, we introduce a multi-channel hyperbolic scoring mechanism that amplifies watermark signals while suppressing noise from weakly aligned candidates. We further propose a diversity-aware filtering strategy that combines hard filtering with soft regularization, extending beyond simple n-gram repetition filters to address semantic redundancy. Experimental results show that SAMark achieves up to 90.2% TP@FP1% under typical paragraph-level paraphrasing attacks, outperforming the strongest prior baseline by more than 30% on average, while maintaining generation quality competitive with unwatermarked text and breaking the robustness-quality trade-off that limits prior methods.

07.
arXiv (CS.CL) 2026-06-16

Semantic-Preserving Prompt Hijacking: A Black-Box Adversarial Attack on Auto-Prompt Optimization

LLMs increasingly integrate auto-suggestion optimization modules, enabling them to rewrite and display user input before generating the final response. While this design aims to enhance transparency and trust, its process of autonomously selecting a single best result from multiple candidate solutions allows attackers to hijack this optimization process by inducing subtle, imperceptible semantic shifts. To address this, we propose a semantic preservation hijacking attack method based on black-box conditions: Adaptive Greedy Local Search. This method hierarchically decomposes the input text, masks key language units, and dynamically adjusts candidate replacement words at predefined semantic checkpoints. This maximizes the deviation between the model output and the original intent while strictly maintaining semantic similarity to the original text. Experimental results on commercial and open-source LLMs demonstrate that, under the same semantic similarity constraints, this method achieves a higher attack success rate than existing attack methods in over 2400 test cases. Code is available at: https://github.com/franz-chang/DOBS

08.
arXiv (CS.AI) 2026-06-11

Categorical Prior Lock-in: Why In-Context Learning Fails for Structured Data

arXiv:2606.11961v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as conditional generators for structured data, relying on in-context learning (ICL) to adapt to new distributions without parameter updates. We investigate the limits of ICL for structured generation under distribution mismatch, using high-cardinality tabular data as a controlled test case, and identify a structural failure mode we term categorical prior lock-in: the inability of ICL to update the model's prior over token distributions inherited from pre-training. Across two 7B-parameter open-weight models, ICL improves numerical fidelity with additional examples but exhibits a sharp ceiling on categorical distributions, failing to reproduce rare classes entirely. Parameter-efficient fine-tuning (LoRA) overcomes these limitations but introduces measurable memorization risk and, in some cases, destabilizes structured output generation, highlighting a fundamental trade-off between adaptability and privacy.

09.
medRxiv (Medicine) 2026-06-19

A soluble bi-specific fusion protein for the improved expansion of human CD8+ CAR-T cells

The success of Chimeric Antigen Receptor (CAR) T cell therapy is heavily dependent on the quality of the final cellular product. Current expansion protocols often rely on reagents that require removal from cell culture media, posing logistical challenges in manufacturing, and can also lead to terminal differentiation. Here, we evaluate the use of a soluble, bead-free T cell activator, T cell expansion protein (T-CEP), as a streamlined alternative for generating potent CAR-T cells. Human T cells were activated with T-CEP or known T cell activators (Dynabeads and TransAct) and transduced with either CD19 or interleukin-13 (IL-13) mutein (tetravariant-13; TV-13)-based CAR lentiviral vectors. Our results demonstrate that T-CEP supports robust CAR-T cell expansion and achieves transduction efficiencies comparable to commercial reagents for both types of CAR-T cells. Notably, T-CEP significantly favored the expansion of CD8+ T cells, yielding an enhanced CD27+ phenotype and a lower CD4:CD8 ratio compared to TransAct. Cytotoxicity assays confirmed that T-CEP-expanded CAR-T cells possess cytolytic function equivalent to commercial reagents for both CARs, while exhibiting lower levels of inflammatory cytokine secretion. In summary, T-CEP represents a competitive alternative to existing expansion agents, as it does not require its removal during CAR-T manufacturing and generates a CD8+ dominant, less-differentiated phenotype without compromising efficacy.

10.
arXiv (CS.CL) 2026-06-18

EARS: Explanatory Abstention for Reliable Sub-Agent Modeling in Large-scale Multi-Agent Systems

In large-scale enterprise settings, centralized multi-agent systems (MAS) are increasingly adopted, in which a coordinator delegates user requests to lightweight, domain-specialized sub-agents. While this architecture improves modularity, scalability, and cost efficiency, its reliability depends not only on accurate routing but also on sub-agents' ability to calibrate their responses to capability constraints. In particular, sub-agents built on smaller fine-tuned models often struggle with such calibration, leading them to over-answer ambiguous, underspecified, misrouted, or unsupported requests and produce hallucinated outputs instead of actionable feedback. To address this challenge, we present EARS (Explanatory Abstention for Reliable Sub-Agent Modeling), a production-oriented framework that reframes sub-agent abstention as an inter-agent communication protocol: a sub-agent does not merely abstain, but exposes an actionable failure state to the coordinator. EARS curates human-agent interaction data using an ensemble of calibrated LLM-as-a-Judge models, producing structured abstention labels and rationales under a taxonomy of sub-agent failure modes. These data are used to fine-tune sub-agents to detect failure conditions and return rationales for coordinator-level clarification, rerouting, or fallback. We evaluate EARS in a large-scale production e-commerce assistant supporting enterprise business intelligence workflows. EARS improves the overall response pass rate from 68.5% to 78.9%, demonstrating that sub-agent-side explanatory abstention improves MAS reliability.

11.
arXiv (CS.AI) 2026-06-12

HarnessBridge: Learnable Bidirectional Controller for LLM Agent Harness

arXiv:2606.12882v1 Announce Type: new Abstract: Large language models are increasingly deployed as agents for long-horizon tasks, yet their performance is shaped not only by model capability and environment design, but also by the harness that mediates agent–environment interaction. Existing harnesses are largely manually engineered, making them difficult to scale as trajectories grow longer and interactions become more complex. In this work, we ask whether harness can be generated by a learnable plug-in module that can be trained in an end-to-end fashion. We introduce HarnessBridge, a lightweight learnable harness controller that parameterizes the agent–environment interface as a bidirectional projection. HarnessBridge learns two bidirectional projections: observation projection, which distills raw trajectories into compact, decision-relevant states, and action projection, which converts proposed actions into executable transitions or trajectory-grounded rejections. We train HarnessBridge on a harness supervision dataset via unified instruction tuning. On Terminal-Bench~2.0 and SWE-bench Verified, HarnessBridge matches or surpasses strong specialized harnesses while substantially reducing token usage and trajectory length, and generalizes from smaller generators to larger commercial models.

12.
arXiv (CS.CV) 2026-06-15

SED:Lightweight Saliency prediction for Event-based data via Distillation

Event-based saliency prediction has gained attention recently, as combining event cameras with saliency estimation can act as an upstream stage that naturally improves the efficiency of downstream eventbased perception at the edge. However, current approaches are either neuromorphic, underperforming on event-based saliency benchmarks, or too heavy for resource-constrained edge applications due to their reliance on transformers or 3D convolutions. Drawing inspiration from efficient convolutional modules, SED and aiming to exploit the temporal information in event data, we propose a lightweight network, trained through knowledge distillation, built on a Depthwise Spatio-Temporal Block (DSTconv) – a factorization of the 3D depthwise separable convolution. Relative to its teacher, our model reduces the model size from 180 MB to 0.32 MB (562x) and the parameter count from 45M to 81k (554x), while matching or outperforming it on the N-DHF1K and N-UCF Sports datasets. Moreover, it generalizes strongly beyond its training distribution, transferring from synthetic to real event data where a model trained from scratch fails.

13.
arXiv (CS.CL) 2026-06-17

Beyond Domains: Reusing Web Skills via Transferable Interaction Patterns

Large language model (LLM) web agents are usually deployed as tool callers: each turn, the model reads a fresh page observation and emits one structured tool action. When every action is a low-level primitive, horizons grow quickly and so do policy-facing LLM completions, dominating latency and cost on benchmarks such as Mind2Web and WebArena. Recent systems therefore wrap repeated interaction fragments as web skills: callable tools built from successful trajectories or induced programs, so one call can replace several primitives. However, prior skill libraries are still triggered mainly by instruction similarity or coarse site metadata, which yields low skill reuse on held-out sites and leaves much of the potential step and token reduction on the table. We present SkillMigrator, an agent that learns reusable web skills and transfers them across sites by matching layout structure rather than specific element references. Each induced skill is stored as a transferable interaction pattern (TIP): the skill paired with a structural sketch of the snapshot at induction time. At test time, SkillMigrator retrieves TIPs by layout similarity and grounds their references on the live page. The rest of the stack is standard: accessibility-snapshot observations with stable references, and fixed tool calling over primitives plus skill invocations. Compared with the state-of-the-art approaches, SkillMigrator reduces the average LLM-action count on successful trajectories by 8-10% across both WebArena and Mind2Web at matched success rate.

14.
arXiv (CS.AI) 2026-06-17

Querying an astronomical database using large language models: the ALeRCE text-to-SQL system

arXiv:2606.18108v1 Announce Type: cross Abstract: We develop a text-to-SQL (structured query language) system based on large language models (LLMs) using in-context learning and apply it to the Automatic Learning for the Rapid Classification of Events (ALeRCE) astronomical database. ALeRCE is a community broker for the Zwicky Transient Facility and the Vera C. Rubin Observatory. The system enables users to query the database in natural language (NL) and generates executable SQL queries. To develop and evaluate the system, we constructed a dataset of 110 NL/SQL pairs. We propose a step-by-step generation framework comprising four modules: schema linking, query classification, prompt decomposition, and self-correction. The performance of thirteen LLMs is evaluated using in-context learning and prompt engineering techniques. Text-to-SQL performance is assessed using the perfect-match (PM) rate for row identifiers (e.g., object identifiers) and column identifiers (i.e., column names). The proposed step-by-step framework consistently outperforms a direct-inference baseline, while the self-correction module consistently reduces execution errors. For Claude Opus 4.6, PM performance on row (column) identifiers is high for simple queries, reaching 0.97 (0.94), and decreases with query complexity to 0.44 (0.72) for medium queries and 0.59 (0.49) for hard queries. Among the thirteen evaluated models, the best-performing LLMs for the text-to-SQL task are Claude Opus 4.6, Gemini 2.5 Pro, Gemini 3 Flash, and GPT-5.2-Codex.

15.
arXiv (CS.CL) 2026-06-16

AthDGC: An Open Diachronic Greek Treebank with Indo-European Parallels

AthDGC ("Athens-PROIEL") is an open, end-to-end workflow and dataset. It is, to the best of our knowledge, the first openly licensed dependency-parsed treebank of Greek that spans eight diachronic periods, namely Archaic, Classical, Koine, Late Antique, Byzantine, Late Byzantine, Early Modern, and Modern Greek, under a single PROIEL XML 2.0 schema, with verse-level cross-alignment of the New Testament to Latin (Vulgate), Gothic (Wulfila), Old Church Slavonic (Marianus), and Classical Armenian. AthDGC builds on the PROIEL Treebank Family (Haug and Johndal 2008; Eckhoff et al. 2018), which established the schema and the Koine-Greek reference set for the project. Annotation uses the Stanford Stanza PROIEL-trained workflow; sentence-level alignment uses LaBSE, a multilingual sentence-embedding model; word-level alignment uses multilingual-BERT attention through the AwesomeAlign procedure. The v0.4 release provides curated samples and the open-source toolkit; the full annotated corpus partitions remain under v0.5 audit on the Greek national HPC. Quantitative scale, per-witness verse counts, and per-period annotated-row counts are reported in the v0.5 release notes, after the audit pass completes. Concept DOI: 10.5281/zenodo.20439182.

16.
arXiv (CS.LG) 2026-06-12

Learning with Simulators: No Regret in a Computationally Bounded World

arXiv:2606.13576v1 Announce Type: new Abstract: Understanding the minimal assumptions necessary for generalization is the fundamental question in learning theory. Unfortunately, most results rely heavily on independence (or some proxy thereof) of the data-generating process, while results for strongly dependent data are far more limited. Towards addressing this gap, we introduce the framework of simulatable processes, where the learner has access to a simulator that approximates the distribution generating the data (which may be an arbitrarily complex and dependent process). Surprisingly, given access to such a simulator, we show that we can recover the same learning guarantees as in the classical setting with independent data, namely, error bounds that depend on the VC dimension. Further, we use this framework to study the power of conditional sampling and show strict statistical and computational advantages in this setting. As a highlight of our framework, we exhibit a single algorithm that simultaneously learns any given VC class under all processes samplable in bounded polynomial time, with regret controlled by the time-bounded Kolmogorov complexity of the process. This provides a significant conceptual broadening of the classical PAC model.

17.
arXiv (CS.AI) 2026-06-19

Too long; didn't solve

arXiv:2604.07593v2 Announce Type: replace Abstract: Mathematical benchmarks consisting of a range of mathematics problems are widely used to evaluate the reasoning abilities of large language models, yet little is known about how their structural properties influence model behaviour. In this work, we investigate two structural length variables, prompt length and solution length, and analyse how they relate to model performance on a newly constructed adversarial dataset of expert-authored mathematics problems. We find that both prompt and solution lengths correlate positively with increased model failure across models. We also include a secondary, exploratory analysis of cross-model disagreement. Under a difficulty-adjusted normalised analysis, both variables retain weak negative associations with realised model separation, slightly stronger for prompt length. Overall, our main robust finding is that structural length is linked to empirical difficulty in this dataset.

18.
arXiv (math.PR) 2026-06-16

A tree-free approach to 3D Yang-Mills Langevin dynamic. Analytic estimates and the existence of a model for a regularity structure

arXiv:2605.14616v2 Announce Type: replace Abstract: Using the multi-index approach to regularity structures due to F. Otto et al., we construct a regularity structure and a model for it associated to the stochastic Langevin equation for the 3D Euclidean Yang-Mills functional. For the model we also obtain global stochastic and global pointwise weighted Besov type estimates which hold almost surely. The model is defined as a limit of a sequence of smooth models introduced with the help of a mollified noise. When the mollification is removed the sequence converges in a certain topology defined with the help of the stochastic estimates. To obtain these results we develop the multi-index approach for systems of equations with vector-valued white noises. This project is motivated by the problem for constructing 3D Euclidean Yang-Mills measure and by the earlier results of the author on the related problem of canonical quantization of the Yang-Mills field on the Minkowski space.

19.
arXiv (CS.CV) 2026-06-15

VideoWeave: Unlocking Geometric Consistency in Video Generation via Joint Geometry-Video Modeling

Large-scale video diffusion models often fail to preserve 3D structure over time, causing geometric drift and implausible motion under viewpoint changes. Existing methods usually enforce geometric consistency by using explicit geometry reconstructions, such as depth maps, point clouds, or reconstructed 3D structures, to define conditions, supervision, or reward signals, making the generator sensitive to errors from upstream geometry pipelines. We propose VideoWeave, a latent-space post-training framework that uses implicit geometry-model features to constrain the generative distribution, providing a more flexible and non-rigid form of guidance that mitigates the impact of reconstruction errors from geometry models. Specifically, VideoWeave adapts these features into geometry latents and jointly models them with video latents in a shared denoising space, allowing geometry to shape the generative distribution during training. To support this process, we build GeoVid-80K, an 80K-video dataset with paired appearance and geometry representations. Experiments on text-to-video and image-to-video generation show that VideoWeave improves geometric coherence while preserving strong visual quality. VideoWeave project page at https://videoweave.github.io/

20.
arXiv (CS.CL) 2026-06-16

When the Chain of Thought Knows Better: Failure Modes in Multi-Turn Reasoning Models

Failures in multi-turn reasoning models are largely invisible to terminal-score evaluation. A model can lock onto an unsafe stance early in a long dialogue, yet its final-turn refusal rate may appear indistinguishable from a robustly aligned baseline. To expose these hidden temporal dynamics, we propose a trace-level diagnostic - the CoT-Output 2x2 safety matrix. This framework labels every turn along two independent axes (internal reasoning and visible output), yielding four operationally defined failure cells: robust alignment, alignment faking, overt jailbreak, and a distinct failure mode we term context-injection failure (where the CoT maintains safe reasoning, but the visible output produces harm, highlighting a multi-turn manifestation of reasoning unfaithfulness). We evaluate three distilled reasoning targets against a fixed attacker across five oversight conditions, collecting 6750 turn-level observations on the Information-Hazard scenario. Our analysis reveals two reproducible vulnerabilities: an oversight paradox where explicit monitoring cues paradoxically increase alignment-faking rates rather than suppress them, and a context-injection failure where models lock onto unsafe external outputs despite safe internal states. We release the full dataset of multi-turn dialogues and CoT traces to support follow-up trace-diagnostic research.

21.
arXiv (CS.AI) 2026-06-12

Proprioceptive-visual correspondence enables self-other distinction in humanoid robots

arXiv:2606.13222v1 Announce Type: cross Abstract: Distinguishing self from others is a prerequisite for social intelligence, yet humanoid robots that increasingly share workspaces with humans still lack this ability. Here we show that a humanoid robot can learn self-other distinction from proprioceptive-visual correspondence, without any identity labels or kinematic models. Once established, this distinction bootstraps a predictive self-model that maps joint configurations to three-dimensional body occupancy, capturing how the robot's body changes with action. In multi-agent scenes involving humans or morphologically identical robots, the system reliably identifies itself, learns a 3D self-model, and supports downstream tasks including target reaching, collision-aware motion planning, and human-to-robot motion retargeting. Together, these results outline a route toward bodily self-representation in robots that act and coordinate alongside others in shared physical environments. Project page: https://euron-zc.github.io/humanoid-self-model/.

22.
arXiv (CS.AI) 2026-06-16

SorryDB: Can AI Provers Complete Real-World Lean Theorems?

arXiv:2603.02668v2 Announce Type: replace Abstract: We present SorryDB, a dynamically-updating benchmark of open Lean tasks drawn from 78 real world formalization projects on GitHub. Unlike existing static benchmarks, often composed of competition problems, hillclimbing the SorryDB benchmark will yield tools that are aligned to the community needs, more usable by mathematicians, and more capable of understanding complex dependencies. Moreover, by providing a continuously updated stream of tasks, SorryDB mitigates test-set contamination and offers a robust metric for an agent's ability to contribute to novel formal mathematics projects. We evaluate a collection of approaches, including generalist large language models, agentic approaches, and specialized symbolic provers, over a selected snapshot of 1000 tasks from SorryDB. We show that current approaches are complementary: even though an agentic approach based on Gemini Flash is the most performant, it is not strictly better than other off-the-shelf large-language models, specialized provers, or even a curated list of Lean tactics.

23.
arXiv (CS.AI) 2026-06-11

The Algorithm Is Not the Behavior: Learned Priors Override Look-Ahead in a Chess-Playing Neural Network

arXiv:2508.21380v3 Announce Type: replace-cross Abstract: Recent mechanistic work has uncovered learned algorithms within neural networks, from modular arithmetic to search and planning in game-playing agents. But does algorithmic structure guarantee algorithmic behavior? We investigate this in Leela Chess Zero, the strongest neural chess engine, where prior work identified learned look-ahead. By extending the logit lens to its move-selecting policy network, we discover that correct puzzle solutions-including immediate checkmates-often appear in intermediate layers but are systematically overridden in the final output, a phenomenon we term "forgotten puzzles". Replicating prior analyses on these positions, we find that look-ahead operates normally-future moves of the correct continuation are represented, causally important, and linearly decodable-ruling out a failure of the algorithm itself. Instead, late layers increasingly shift toward prioritizing safe play over aggression. To test whether this shift drives the override, we steer the model against these preferences and recover 61.7% of forgotten puzzles, providing causal evidence that safety priors override algorithmically computed solutions. These findings demonstrate that algorithmic structure does not guarantee algorithmic behavior: a model can internally solve a problem and still output the wrong answer.

24.
arXiv (CS.CL) 2026-06-16

DRA-GRPO: Your GRPO Needs to Know Diverse Reasoning Paths for Mathematical Reasoning

Post-training LLMs with Reinforcement Learning, specifically Group Relative Policy Optimization (GRPO), has emerged as a paradigm for enhancing mathematical reasoning. However, standard GRPO relies on scalar correctness rewards that are often non-injective with respect to semantic content: distinct reasoning paths receive identical rewards. This leads to a Diversity-Quality Inconsistency, where the policy collapses into a narrow set of dominant modes while ignoring equally valid but structurally novel strategies. To bridge this gap, we propose Diversity-aware Reward Adjustment (DRA), a theoretically grounded framework that calibrates the reward signal using the semantic density of sampled groups. By leveraging Submodular Mutual Information (SMI), DRA implements an Inverse Propensity Scoring (IPS) mechanism that effectively de-biases the gradient estimation. This creates a repulsive force against redundancy, driving the policy to achieve better coverage of the high-reward landscape. Our method is plug-and-play and integrates seamlessly with GRPO variants. Empirical evaluations on five math benchmarks demonstrate that DRA-GRPO consistently outperforms strong baselines, achieving an average accuracy of 58.2% on DeepSeek-R1-Distill-Qwen-1.5B with only 7,000 training samples and $55 cost, highlighting the critical role of diversity calibration in data-efficient alignment. The code is available at https://github.com/xiwenc1/DRA-GRPO.

25.
arXiv (CS.CV) 2026-06-16

Discriminative Span as a Predictor of Synthetic Data Utility via Classifier Reconstruction

In many real-world computer vision applications, including medical imaging and industrial inspection, binary classification tasks are characterized by a severe scarcity of positive samples. A widely adopted solution is to generate synthetic positive data using image-to-image transformations applied to negative samples. However, a fundamental challenge remains: how can we reliably assess whether such synthetic data will improve downstream model performance? In this work, we propose a geometry-driven metric that predicts the utility of synthetic data without requiring model training. Our approach operates in the embedding space of a pre-trained foundation model and represents the dataset through difference vectors between samples. We evaluate whether the weight vector of a linear classifier can be expressed within the subspace spanned by these variations by measuring the relative projection error. Intuitively, if the variations induced by synthetic data capture task-relevant directions, their span can approximate the classifier, resulting in low projection error. Conversely, poor synthetic data fails to span these directions, leading to higher error. Across multiple datasets and architectures, we show that this metric exhibits strong correlation with downstream classification performance of CNNs trained on mixtures of real negative and synthetic positive data. These findings suggest that the proposed metric serves as a practical and informative tool for evaluating synthetic data quality in data-scarce settings.