Research

• General in-context learning for embodied agents1 paper raises it · 0 address it

  Improving adaptability of models to new tasks and environments without task-specific fine-tuning is critical for scalable embodied intelligence.

  • Has GPT-6-Astra Achieved Embodied Intelligence?raises
• Cost and latency in agentic systems2 papers raise it · 1 addresses it

  High computational and time costs limit practical deployment of agentic systems for real-world tasks.

  • Has GPT-6-Astra Achieved Embodied Intelligence?raises
  • PremiseCheck: Efficient Fact-Checking of Human and AI Peer Reviewsraises · addresses
• Physical-state verification and failure recovery1 paper raises it · 0 address it

  Robustness to irreversible failures and accurate state tracking are essential for reliable embodied control.

  • Has GPT-6-Astra Achieved Embodied Intelligence?raises
• Evidence-grounded multimodal reasoning benchmarks1 paper raises it · 1 addresses it

  Current benchmarks often fail to ensure models use necessary audio-visual evidence, weakening evaluation reliability.

  • EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual Reasoningraises · addresses
• Length generalization in sequential state tracking1 paper raises it · 1 addresses it

  Models struggle to maintain performance on problems much longer than their training data, limiting scalability.

  • PAR: Progress-Aligned Recurrence for Length Generalization in State Trackingraises · addresses

Named by an AI from these papers' abstracts.

• Training & fine-tuning42+90.9%
• AI safety29+61.1%
• Computer vision29+31.8%
• Evals60+25.0%
• Federated learning50+22.0%
• Code models13+18.2%
• Quantum machine learning26+13.0%
• Image & video generation20+5.3%

• arXiv cs.CLOct 15

  Trajectory-Guided Steering: Mitigating Spurious Feature Bias in Clinical Classification

  Fardin Ahsan Sakib, Ziwei Zhu, Ozlem Uzuner

  Large language models rely on spurious features in clinical classification: swapping a single gender token in clinical notes alters gender-irrelevant diagnoses in up to 53.5% of cases. Existing inference-time corrections either need counterfactual data built around the specific spurious feature or behave unpredictably. We observe that how a model's preference between answers evolves across layers, its prediction trajectory, indicates for each document whether a correction is likely to help. We propose Trajectory-Guided Steering (TGS), which decides from the trajectory alone whether to correct each prediction and in which direction, with no counterfactual data and no knowledge of the spurious feature. On 16 clinical classification tasks from i2b2, TGS fixes 17.7% of errors while breaking only 1.3% of correct predictions and stays net-positive across all tested hyperparameter settings; all baselines are net-negative or near zero. The same corrections reduce reliance on the spurious feature: on the most affected diseases, gender-swap inconsistency falls sharply (hypertension 53.5% -> 38.0%, diabetes 35.1% -> 21.5%). Applying the correction to every document instead raises breaks on…

  PDF
• arXiv cs.CLOct 15

  When Does Uncertainty Measure RAG Hallucination Risk?

  Rajeel Ahmad Ansari, Tautik Agrahari

  Uncertainty can measure hallucination risk in retrieval-augmented generation (RAG), but only when retrieved evidence constrains the answer. When evidence is present, unsupported factual claims become less likely; when retrieval coverage is poor, the model can answer from internal knowledge and assign high probability to unsupported claims. We test this across four instruction-tuned models on HotpotQA and SQuAD. After matching faithful and hallucinated answers within word-count bins, length-only classifiers fall near chance, while perplexity reaches 0.81--0.91 AUROC on targeted unsupported edits. Varying retrieval depth shows why the score changes meaning: detection can weaken with only the top-ranked passage, then recover after adding a second passage. A combined uncertainty and semantic-drift score improves calibration, and measurements remain above chance on RAGTruth QA. Results support a conditional view of uncertainty as a useful measure of RAG hallucination risk.

  PDF
• arXiv cs.CVOct 15

  Shift-Reliance Alignment: Ranking Vision-Language Models by the Cues They Trust

  Faisal Mohammad, Mohamed Younis

  Vision-language models (VLMs) are usually chosen from leaderboards, which give one ranking for every application. On questions about objects that a scene suggests but does not contain, however, the model with the lowest error on the POPE hallucination benchmark is not the one with the lowest error on such questions. For yes/no questions about object presence, we show that the best model for a target domain can be predicted without running any candidate on that domain, given 650 to 1,200 labeled target items that calibrate a detector. We measure each model's reliance on a cue, a feature such as scene context or text in the image, by editing 100 labeled source images so that the cue suggests the wrong answer, and comparing with a matched control edit. Reliance predicts which VLMs are misled by physical sticker attacks (Spearman .93 on nine models not used to design the score) and which hallucinate on unedited COCO photographs, beyond what their tendency to answer yes explains. Shift-Reliance Alignment (SRA) scores a model by its source error on the labeled source questions plus its reliance multiplied by the shift. The shift is the proportion of target questions on which the cue…

  PDF
• arXiv cs.CVOct 15

  ProactGen: Learning Proactive Clarification for Video Generation

  Zhixiao Wang, Yang Chen, Yifan Xu, Chenbin Li, Yue Chang +3

  Recent advances in video generation let users create high-quality videos from natural language descriptions, yet obtaining results that match user intent remains time-consuming: individual generations are computationally expensive, and underspecified prompts force repeated cycles of generation, inspection, and revision. Existing video prompt optimization methods improve generation through automatic rewriting, but cannot supply the ideas and preferences that were not in the initial description. We propose ProactGen, a video generation framework that learns proactive clarification, eliciting user intent through questions grounded in generated video drafts. Specifically, we train a vision-language model as a questioning policy with reinforcement learning: observing the initial prompt and a video draft, it learns to ask the questions most valuable to downstream generation, rewarded by the alignment between the final generated video and a target reference. We further find that effective questioning does not require drafts of final-output quality, and exploit this property by generating drafts with fewer denoising steps, keeping the added cost of clarification low. Experiments show…

  PDF
• arXiv cs.CVOct 15

  Three LiDARs, Twenty-Five Datasets: Cross-Source Generalization in Tree-Species Recognition

  Jae Joong Lee, Bedrich Benes

  FOR-species20K is the largest public LiDAR benchmark for tree-species classification, pooling 25 independently collected terrestrial, mobile, and airborne laser-scanning datasets across 33 species. We introduce a multi-resolution, multi-view ConvNeXtV2 model that improves official hidden-test accuracy from 79.5% to 85.0%, ranking first at the time of submission. We then examine what this score measures. Because the hidden test is species-balanced, its overall accuracy is more comparable to a macro metric than to accuracy under the public set's long-tailed distribution. Moreover, the official random split places every source dataset in both training and evaluation. A logistic probe recovers source identity from out-of-fold model probabilities with 69% accuracy, while geometric features identify the source above 90% within two widely shared species; 18 of the 33 species occur in only one source. Under a source-species-cell holdout, evaluated on 5,566 trees from eight eligible species that retain the same LiDAR-type training support in other sources, the unweighted species mean falls from 82% to 36%, and tree-weighted accuracy falls from 95% to 60%. Count-matched retraining and…

  PDF
• arXiv cs.AIOct 15

  Beyond Explicit Intent: Discovering and Shaping Artistic Intent through Human-AI Co-Creation

  Renye Yan, Jialun Zhong, Jikang Cheng, Yi Gao, Yanzeng Li +1

  Generative AI is becoming an important medium for artistic creation, yet existing interactions often assume that creators can specify intentions in advance and accurately communicate them through text prompts. Artistic intentions, however, may be vague, latent, or recognizable only after seeing concrete visual possibilities. We propose a human-AI co-creation paradigm beyond explicit intent expression. By expanding model exploration, we expose artists to diverse visual possibilities, allowing them to identify previously unarticulated preferences through observation, comparison, and selection. We then integrate artists' curatorial judgments into reinforcement learning, assigning different learning strengths based on preferences over generation trajectories, so recognized directions are more strongly reinforced and the model gradually adapts to individual aesthetics. This creates a continuous "explore-identify-shape-re-explore" loop in which artistic intent is progressively discovered, externalized, and shaped. Through three user studies, we examine latent-preference discovery, personalized preference alignment, and the externalization of vague, difficult-to-express creative…

  PDF
• arXiv cs.LGOct 15

  Rarity of Singletons Matters & Supplementary Materials

  Pragatheeswaran Vipulanandan, Kamal Premaratne, Dilip Sarkar

  Estimating missing-mass and unseen-species counts from finite samples relies heavily on rare-event occupancy statistics, particularly singletons. In small-sample regimes, these statistics can be highly variable, making classical estimators sensitive to sampling fluctuations and potentially unreliable. We propose a fragility-aware (FA) framework that explicitly quantifies this sensitivity and adaptively regularizes conventional missing-mass and unseen-species estimators. Using singleton-based fragility measures derived from probability mass and occupancy counts, we formulate an entropy--Kullback--Leibler (KL) optimization that adjusts the reliance on base estimates according to their fragility. The resulting objectives are strictly concave, admit unique closed-form solutions, and require no Poissonization or asymptotic approximations. The framework provides a lightweight correction to classical estimators, including Good--Turing and Good--Toulmin, while reducing reliance on potentially unstable rare-event statistics. Across 360 experimental configurations spanning synthetic distributions and real-world language corpora, our method remains competitive in stable regimes while…

  PDF
• arXiv cs.AIOct 15

  At the Limits of Computability: Steps Towards Autonomous AI Discovery on the Busy Beaver Frontier

  Marco Dos Santos, Vivek Veeriah, Alex Havrilla, Andrew Bolt, Goran Zuzic +2

  Recent breakthroughs in AI for mathematics---ranging from Olympiad medal performances to resolving long-standing open conjectures---have rapidly expanded what automated systems can achieve. Yet these successes primarily concern predefined, human-formulated problems, leaving open how AI systems can autonomously uncover mathematical structure and formulate new questions. To push beyond static challenges, we explore the Busy Beaver (BB) domain, an uncomputable frontier that is fundamentally open-ended. Because no algorithm can compute the Busy Beaver number $\mathrm{BB}(n)$ for every $n$, this domain provides a natural setting for sustained exploration through mathematical conjecture and proof. In this paper, we introduce a Gemini-based agentic pipeline that automatically discovers accelerations of Turing machines, expressed as compact mathematical rules summarizing many execution steps. Our system integrates pattern recognition, automated conjecturing, informal mathematical reasoning, and formal theorem proving in Lean. With this framework, our system autonomously discovered and formally verified novel Collatz-like accelerations for eleven machines in $\mathrm{BB}(6)$. Subject to…

  PDF
• arXiv cs.CVOct 15▲ 15

  Stuck on Shuffle: On Attribute-Object Binding in Vision-Language Models

  Netanel Y. Tamir, Michael Green, Issar Tzachor, Or Shimshi, Dvir Samuel +2

  Retrieval is a fundamental task in multimodal systems: given an image or text query, the goal is to find the most relevant images or texts in a collection, often using a shared embedding space. Reliable retrieval requires these embeddings to capture not only which concepts are present, but also how they are composed. One key compositional challenge is attribute--object binding: encoding which attributes belong to which objects. Existing benchmarks provide a limited view of this capability, typically covering a subset of modality pairings, simple two-object swaps, or small attribute sets. Consequently, it remains unclear how robustly current models preserve attribute--object bindings across different modality pairings and more complex settings. We therefore introduce SABER, a controlled benchmark for evaluating attribute--object binding across all four image and text modality pairings, spanning diverse attribute types and scenes with up to ten objects. Evaluating models on SABER reveals substantial binding failures in both CLIP-style dual encoders and recent MLLM-based embedders. These findings motivate us to ask how binding can be improved efficiently. While multimodal…

  Code
• arXiv cs.LGOct 15

  Can LLMs predict the outcomes of machine learning experiments?

  Ayush Bansal, Aldo Pareja, Sean Klein, Serge Didenko, Artem Lukoianov

  Recent advances in AI make Recursive Self-Improvement (RSI) increasingly plausible. Large Language Models (LLMs) can already propose, select, and run Machine Learning (ML) experiments. However, the space of possible design choices in ML is vast. An LLM that proposes an experiment needs a reliable intuition of whether it is worth running given the timing constraints of long training runs. In this work, we measure such intuition in current frontier LLMs by evaluating how accurately they predict the impact of self-generated ML experiment ideas before these are put into training runs. To this end, we construct a dataset of 288 LLM-proposed ideas on eight MLE-bench tasks. We then prompt five frontier LLMs (GPT-6 Astra, Claude Opus 5.5, Claude Fable 5.1, Kimi K3, and GLM-5.3) to predict the impact of each idea and compare their predictions with the actual outcomes after training. We report four main findings. First, LLM predictions are substantially better than random chance but far from accurate: even the best model, Claude Opus 5.5, explains only half of the variance in outcomes (R^2 = 0.49). Second, we notice that all tested models systematically underestimate the magnitude of the…

  PDF
• arXiv cs.CLOct 15

  Broad and Efficient PII Detection with Unified Labels

  Andy Hsu, Konstantin Berlin

  Personally Identifiable Information (PII) detection models must efficiently find identifiers across diverse contexts while maintaining a low false-positive rate (FPR) on high-volume traffic. Recent detectors increasingly rely on complex architectures that raise deployment cost, yet many are fine-tuned on only a few synthetic datasets that limit their coverage. Combining public datasets is challenging because their taxonomies differ, and even after categories are mapped to a common schema, labels can disagree on what counts as a detection, producing inconsistent training and evaluation signals. To address these problems, we present a practical 141M-parameter encoder-based detector trained on 1.12 million examples from 77 public datasets, relabeled under a detailed definition of broad contextual detection for 12 structured PII categories. Evaluated on the AI4Privacy and Nemotron-PII benchmarks at a low-FPR operating point, our model achieves the highest recall against larger baselines in 11 of 14 comparisons under our labels and 9 of 14 under the original labels. Its largest recall gains include useful additional coverage, such as multilingual disclosures and IP addresses embedded…

  PDF
• arXiv cs.ROOct 15

  Weakly Supervised Discovery of Motion Styles for Compositional Robot Control

  Heedo Woo, Antonio Marino, Eduardo Sebastián, Amanda Prorok

  When a robot possesses more degrees of freedom than a task requires, multiple trajectories can achieve the same objective using distinct motion styles. Compositional robot control aims to preserve a selected style while simultaneously executing an independent task command, effectively balancing these potentially competing objectives. In this paper, we present CODES, a framework for compositional robot control that discovers a coherent style repertoire from diverse recordings and composes it with independent commands. First, motion retargeting maps the recordings to motions executable by the target robot; next, weakly supervised consensus encoding and density-based clustering identify styles within and across recordings without a prescribed count; finally, reinforcement learning trains a policy to compose the discovered styles with independent commands. We evaluate CODES on a Unitree Go2 quadruped across two simulators and physical hardware. Compared to baselines, CODES significantly improves kinematic coherence. Furthermore, our controller successfully tracks velocity commands that generalize beyond the support of the recorded data, exhibiting robust and distinguishable style…

  PDF
• arXiv cs.LGOct 15

  Diagnosing GNN Failure Modes with Layer-Wise Probe Entropy

  Shanzi Bao, Mar Gonzàlez I Català, Pietro Liò

  Graph neural networks (GNNs) are typically evaluated only at their final layer, obscuring how task-relevant information evolves during message passing. We introduce a post-hoc diagnostic that freezes a trained GNN, fits an independent linear probe to each layer, and tracks probe accuracy alongside normalized predictive entropy. Across real and synthetic graph settings, layer-wise probe entropy distinguishes beneficial, neutral, and harmful smoothing, a capability not demonstrated by existing geometric measures. Moreover, by comparing shallow and deep models, we show that the onset of entropy degradation separates the two main GNN failure modes: structural mismatch under heterophily, which appears from the earliest layers even in shallow models, and excessive-depth oversmoothing, which emerges only as propagation depth increases. These results establish probe-entropy trajectories as a tool for analyzing information flow in GNNs and motivate future work on adaptive depth selection and early-exit mechanisms.

  PDF
• arXiv cs.CLOct 15

  The Model Knows Better: Defending Medical Multi-Agent Systems Against Prompt Injection

  Fardin Ahsan Sakib, Ziwei Zhu, Ozlem Uzuner

  Medical multi-agent systems, in which LLM agents discuss a case, are increasingly studied for clinical decision support. Prompt injection, text that an attacker adds to a model's input, can manipulate LLMs into wrong decisions. How it affects these systems, and how to protect them, remains unclear. Across two systems, two model families and three benchmarks, a single injected passage succeeds in up to 61% of cases, and collaboration does not consistently protect against it. Existing defenses help only partly: they miss injections that look like ordinary clinical text. Yet when asked to check the injected text against the patient's findings, a model that followed the injection usually rejects it. We therefore propose a training-free defense that verifies the clinical judgments behind the system's decision. Following clinical practice, it separates the patient's findings from clinical judgments, such as diagnoses, recommendations and instructions, and checks each judgment against the findings alone, so the decision rests only on supported judgments. Our defense outperforms all baselines on every benchmark, lowering attack success on MedQA to 4-7% while maintaining accuracy on cases…

  PDF
• arXiv cs.LGOct 15

  Flat Loss, Evolving Features: Grokking in Modular Addition

  Akash Kumar, Karthik Sankaranarayanan

  Can a network keep learning while its training cross-entropy stays nearly flat at a positive level? We show that it can. On modular addition, a standard testbed for grokking, ReLU networks trained with supported-Polar hidden updates, normalized output-layer updates and weight decay can keep 100% training accuracy while their hidden features change and held-out accuracy rises from zero to above 90%, at relatively flat training cross-entropy. Correct classification need not minimize cross-entropy, which still depends on prediction confidence. We prove this coexistence for one-hidden-layer networks in two finite-window theorems from constructed initializations: one gives a constant gain in the Fourier fraction of the hidden features, and the other holds at arbitrarily low positive loss, where a readout refitted on the training data generalizes before the network. Experiments from random initialization show the same coexistence on a window fixed in advance, and in deeper networks held-out accuracy also rises over post-hoc windows with a net increase in training cross-entropy. With a last-layer rate set per training fraction and head decay scaled with the prime, deeper networks at…

  PDF
• arXiv cs.CLOct 15

  SAG2: Embedding-Free Multi-Hop Retrieval with BM25 and JEV

  Yuchao Wu

  Strong multi-hop retrieval should be practical to deploy without expensive corpus preparation or complex database infrastructure. We present SAG2, an embedding-free retrieval architecture built around a single SQLite database, BM25, and contextual evidence decisions from JEV. SAG2 follows the query-time association principle of SAG (SQL-Retrieval Augmented Generation), while replacing its event-entity and dense indexes with a lexical index over original passages. Selected passages supply terms for further retrieval, and newly retrieved evidence guides subsequent decisions. This moves semantic evidence selection to query time, eliminating offline embedding computation, LLM-based event-entity extraction, and dedicated database servers. The resulting system combines lightweight preparation with strong retrieval accuracy: on 1,000-question subsets of MuSiQue, 2WikiMultiHopQA, and HotpotQA, SAG2 achieves the highest Recall@5 and Recall@10 among the compared methods, with Recall@10 of 86.62%, 97.43%, and 99.00%. It also reaches 81.10% document Recall@10 over the 511,962-document EnterpriseRAG-Bench corpus. Indexing all 11,656 MuSiQue passages takes 2.36 seconds, produces a 25.87 MiB…

  PDF
• arXiv cs.CLOct 15

  Mitigating Demographic Disparities in Explanation-Human Alignment via Rationale Regularization

  Suryansh Talukdar, Taha Sarwat, Jonathan Rusert

  We study explanation fairness in hate speech detection. High classification performance does not always mean that a model produces equally plausible explanations for different demographic groups. Using the HateXplain dataset, we introduce Plausibility-EFG, a metric that measures the largest difference in explanation plausibility across demographic groups by comparing model-generated token attributions with human-annotated rationales. We evaluate BERT, RoBERTa, and ADV-RoBERTa, an adversarially trained model, using Integrated Gradients and SHAP. Our results show that ADV-RoBERTa improves classification performance and alignment between explanations and human rationales while reducing explanation disparities across demographic groups. These improvements are observed with both attribution methods and remain consistent across training, validation, and test splits. We also show that reducing prediction-level bias alone does not eliminate disparities in explanations, suggesting that predictive fairness and explanation fairness are related but distinct objectives. Finally, our fairness-performance analysis shows that stronger adversarial regularization can further reduce…

  PDF
• arXiv cs.LGOct 15

  Initiating Removal from Operational Symptoms in Federated Vehicular Networks

  Zihao Ding, Jun Huang, Qiang Duan

  Connected vehicles train a driving model through federated learning (FL) and drive with it, so a faulty contribution often surfaces only as a delayed symptom on the road. Federated unlearning (FU) assumes that a removal request names what to forget. A symptom names no vehicle or round, and aggregation has already spread its cause across the fleet. We propose \underline{A}ggregation-\underline{T}ime \underline{O}perational-evidence \underline{N}eutralization of \underline{E}ffects (\method{}), which makes unlearning a term of the FL aggregation rule. \method{} projects everyday driving errors and reported symptoms onto a subspace spanned by recent global increments. It propagates the surviving effect of every contribution with curvature estimated from these increments and scores all contributions in one backward pass. A two-tier correction suppresses suspected effects in every round and eliminates confirmed ones after a one-round probe. We prove that backward scoring matches forward scoring and that the elimination is exact when training is flat outside the subspace. On three driving datasets, \method{} leaves 43\% to 61\% less residual harm than the strongest baseline in eight of…

  PDF
• arXiv cs.LGOct 15

  Noise-Robust Forced Alignment for Spontaneous Speech

  Laurin Wagner, Mario Zusag

  We show that forced aligners degrade on spontaneous, noisy real-world speech, and that fillers, laughter, and disfluencies corrupt word boundaries, impacting many downstream applications. We revisit the classical GMM-HMM aligner with upsampled WavLM features instead of MFCCs, explicit phone units for fillers and non-lexical vocal events trained on verbatim-annotated spontaneous speech, and a discriminative projection fitted on clean and noise-corrupted feature pairs to improve robustness. The pipeline is self-bootstrapped from 149h of curated audio without manual or external boundary targets. On conversational (Buckeye) and disfluent (FluencyBank) speech it reaches word-boundary MAEs of 20.6 and 76.9ms, 19% and 17% lower than the strongest of ten baselines. Catastrophic misalignments (>200ms) are 2.7 times rarer than Montreal Forced Aligners (MFA) best model even after widening MFA's search beam, which we show removes a large pruning artifact from its conversational error; event units halve the error next to laughter, while being meaningfully more robust than existing aligners.

  PDF
• arXiv cs.ROOct 15

  Embodiment-Classifier Guidance: Transferring Frozen VLAs Across Robot Dynamics without Demonstrations

  Devin Han, Hershraj Niranjani

  Flow-based policies such as $π_0$ are trained on a fixed set of embodiments, and transferring them to new ones is an open problem. To adapt a frozen $π_0$ to a new embodiment without additional demonstrations or parameter updates, we train a lightweight embodiment classifier to distinguish successful action chunks on the source body from those on the target body, using a few calibration tasks. On transfer tasks that $π_0$ saw on its source body but the classifier was never trained on, we nudge each denoising step of $π_0$ with the classifier's gradient, which estimates the direction from the distribution of successful actions on the source body toward that on the target body. On the LIBERO simulation benchmark with four modified controllers, embodiment-classifier guidance raises $π_0$'s average success on the target bodies from 72% to 80% across 12 suite-controller pairs, with a 31% relative reduction in failures (11 of 12 gain; 5 significant after Holm correction). In simulation, guidance is data-efficient, needing few target-body rollouts per task, which makes it promising for sim-to-real transfer, where rollouts are costly. Guidance is mixed on weak wrists and harmful on a…

  PDF