Inherit-MAS introduces explicit inheritance at both workflow and execution levels for multi‑agent systems built from large language models. A meta‑model generates a workflow of specialized agents, while a judge evaluates each candidate and diagnoses deficiencies. During refinement, workflow inheritance discards unhelpful nodes and applies validated edits, and execution inheritance reuses stored results only when the request and context match, reducing redundant computation.
By Songtao Wei, Yi Li, Zhichun Guo, Bingzhe Li
TPBench is a benchmark for dialogue compression that focuses on turning-point eviction, where a user’s change in intent or constraints is lost during compression. It evaluates three information targets—initial user goal, current slot value after revision, and both together—using existing dialogue-state annotations from MultiWOZ and SGD, without requiring new crowdsourcing. The benchmark shows that different compression methods rank differently across probes, and that deleting the turn containing a user update significantly harms current-value accuracy, while removing irrelevant turns does not.
By Minji Park, Seunghyun Yoon, Hyuk Lim
Sentry is a failure‑management layer for large language model agents that learns from failures at test time. It retrieves relevant lessons from an external playbook when a failure occurs, verifies recovery without task rewards, and stores new lessons only if recovery succeeds, keeping the playbook out of the agent’s context. Across multiple benchmarks, Sentry outperforms both runtime‑intervention and context‑evolution baselines, and its lessons transfer to unseen tasks.
By Changxiu Ji, Amy Lu, Qizheng Zhang, Kunle Olukotun
WAMpy is a Python framework designed to efficiently synthesize Prolog programs by compiling Prolog clauses into NumPy array-based WAM instructions and enabling partial recompilation against fixed background knowledge. It accelerates performance-critical routines with Numba JIT compilation, and benchmarks show that WAMpy outperforms SWI-Prolog accessed from Python via Janus in repeated compilation-and-evaluation workloads.
By Dominik Magiera, Lukas R\"ohrig, Frank J\"akel
FrugalEvo introduces a cost‑aware evolutionary framework that pairs a high‑cost, powerful LLM with a cheaper LLM to explore and refine program solutions, respectively. It incorporates a cache‑efficient evolution process that maximizes prefix reuse across steps and defines a Budget‑Aware Area Under the Curve (BA‑AUC) metric to evaluate solution quality within a fixed cost budget. Across ten mathematical and systems optimization tasks, FrugalEvo matches or surpasses state‑of‑the‑art baselines in final solution quality and achieves higher BA‑AUC on nine tasks, while delivering new state‑of‑the‑art performance on circle packing for costs as low as $0.55 to $1.68.
By Hui Chen, Xuan Qi, James Xu Zhao, Zhaopeng Feng, Shilong Liu, Kuang Xu, Pang Wei Koh, Bryan Hooi
The paper proposes a training‑free detector that uses sentence‑level context sensitivity to identify unsupported content in retrieval‑augmented generation (RAG) answers. By re‑scoring each sentence with full context, no context, and each chunk removed, the method flags the chunk whose removal most reduces a sentence’s likelihood as the likely source. Evaluated on RAGTruth, TofuEval, and RAGBench, the detector outperforms answer‑level faithfulness scores, achieving AUCs up to 0.745 and matching per‑chunk fact‑checkers while using only a fraction of the compute required by large‑language‑model judges.
By Mohamed Aly Bouke
The paper investigates how different nearest‑neighbour retrieval methods perform as baselines for predicting molecular fingerprints from MS/MS spectra. It systematically compares several variants that differ in the information assumed available at inference, showing that these assumptions significantly affect performance. The goal is to establish stricter baselines for more rigorous benchmarking and progress measurement in the field.
By Ling Min Serena Khoo
Latent-MOPD is a new on‑policy distillation method that allows a single large language model student to learn from multiple specialist teachers by using both the teachers’ output distributions and their hidden state representations. The approach selects late‑layer targets based on teacher‑student relationships, bridges hidden width differences with a shared projection, and groups updates by domain, enabling gradual shift from representation to token supervision. Experiments show that Latent‑MOPD outperforms token‑only, representation‑only, and uniform‑averaging baselines across nine benchmarks in math, code, and logic, and even surpasses the best individual teacher on most tasks.
By Zhengyu Fang, Seoyeon Hong, Jie Yang, Muyang Li, Koyoshi Shindo, Brandon Joseph Lwowski, Jing Li
The paper introduces Gated Slot Attention-2 (GSA2), a new linear‑attention architecture that combines a Gated Oja Rule‑2 for key‑side memory correction with a Gated Delta Rule‑2 for value‑side correction, both operating through shared latent slots. It also presents a hardware‑efficient chunkwise algorithm for parallel training. Experiments show that GSA2 consistently outperforms strong linear‑attention baselines while maintaining linear‑time sequence modeling and constant‑memory recurrent decoding.
By Ruijie Li, Shengnan Ding, Weimin Zhang, Derick Tang, Zhanpeng Zeng, Qinsong Zeng, Ming Chen, Jiaxi Hu, Yuxuan Liang
PaMIR (Public Arrival-ordered Measurement for Inference in Risk) is an open benchmark for credit‑default prediction that addresses the challenge of scarce and late‑arriving labels. It aggregates 19 public datasets—totaling 1.24 million loans, firms, and card accounts from nine countries—into a single, leakage‑audited recipe that is never redistributed. Models are evaluated via repeated i.i.d. splits and a label‑delayed stream, with AUC reported by label budget, and a synthetic‑data harness tests training rows without exposing held‑out data.
By Mikhail Liashkov, Ilyas Varshavskiy, Shuhratjon Khalilbekov, Azizjon Azimi, Bonu Boboeva
SCAD (Structured Credit Assignment and Distillation) is a new method for training long‑horizon agents that separates planning from bounded subtask execution, distills execution locally, and refines planning credit using cross‑rollout subtask prefix trees. It gives planning full terminal credit while providing execution with positive terminal credit and teacher guidance. On all tested benchmarks, SCAD improves macro‑average accuracy by 4.48 percentage points on text tasks and 4.19 points on multimodal tasks compared to the strongest baseline.
By Shangyang Wu, Shuai Zhao, Ziyue Zhu, Jinyang Wu, Anh Tuan Luu, Haoran Luo
The study benchmarks seven deep tabular generative models against trivial baselines across a size ladder of 8 public datasets (200–20,000 rows) and 4 small clinical datasets, totaling 2,220 runs. Using AUROC of classifiers trained on synthetic data and tested on real data, it finds that in 23 of 24 model–dataset pairs no deep model surpasses the best trivial baseline beyond seed noise, and the best baseline wins 40 of 49 evaluated cells. Rankings of deep models are relatively stable below 5,000 rows, but the top two methods often differ by less than seed variation, and rankings on small clinical datasets only moderately align with those on subsampled large datasets.
By Shivam Shrivastava
Dual-Context Analog Retrieval for Time Series Forecasting (DuoTS) introduces a two-stage forecasting approach that first generates a base forecast and then refines it patch by patch. Each refinement step fuses a current context, which focuses on recent tokens, with a detail context that incorporates retrieved analogs and their subsequent trajectories. Experiments on real-world datasets demonstrate that DuoTS achieves state‑of‑the‑art performance, and ablation studies confirm the importance of both contexts and the refinement mechanism.
By Jung Min Choi, Ngoc Son Le, Ibram Abdelmalak, Vijaya Krishna Yalavarthi, Lars Schmidt-Thieme
Power‑SMC is a training‑free sampling method that targets the sequence‑level power distribution used to sharpen large language model outputs for reasoning tasks. It maintains multiple candidate sequences, assigns importance weights based on how well each matches the power distribution, and periodically prunes low‑scoring candidates to keep inference latency close to standard decoding. Experiments on MATH500, GSM8K, GPQA, and HumanEval show that Power‑SMC matches or surpasses Metropolis–Hastings sampling in accuracy, preserves output diversity, and can accelerate inference speed by up to 17.6×.
By Seyedarmin Azizi, Erfan Baghaei Potraghloo, Minoo Ahmadi, Souvik Kundu, Massoud Pedram
The paper introduces <flowreader>, a method that casts evidence selection for long multimodal documents as a minimum‑cost flow problem over a multimodal content graph. It uses spectral decomposition to identify latent query‑relevant aspects and allocates a fixed evidence budget proportionally to their spectral energy, ensuring aspect coverage without a language‑model planning call. On the VisDoMBench benchmark with Qwen3‑VL‑32B, <flowreader> achieves the highest macro accuracy (68.9%) and outperforms prior systems by 2.7 points, while using fewer content nodes and reader tokens.
By Ambuj Mehrish, Sebastiano Vascon
The paper introduces XiangqiBench, an executable benchmark for evaluating large language model agents in Chinese chess. It tracks 8,568 multi‑turn trajectories from 12 frontier LLMs, revealing that metrics such as the Conversion Gap, Consistency Gap, and Simulation Gap overstate true closed‑loop success. The study shows that simply naming the correct move is insufficient; agents must reliably carry a plan through to a verified outcome against an opponent.
By Yekun Chai, Qiwei Peng, Haoyi Xiong
The paper investigates how post‑softmax attention weights behave across different language models by marginalizing over query positions to create a joint token‑head space. Across more than 60 diverse LLMs, token‑wise marginal attention reveals a text‑intrinsic signal that is conserved across models, while head‑wise marginal attention yields a model‑private signature that is consistent across documents. This insight enables a practical method for estimating per‑head budgets for key‑value cache eviction, achieving competitive performance on standard benchmarks without per‑document recomputation or training.
By Valentino Maiorca, Walter Nelson, Francesco Locatello
ReSCUE is a unified framework for simultaneous sign language translation on unsegmented long‑form videos, aligning training and inference with realistic streaming conditions. It incorporates inference‑aware training to manage partial inputs, non‑signing pauses, and multi‑sentence contexts; stabilized re‑translation for low‑latency, revisable predictions with reduced output flicker; and a sentence commitment mechanism for online segmentation and memory management. Experiments show ReSCUE achieves lower latency and superior translation quality in low‑latency settings, approaching oracle offline systems on long‑form datasets while operating at substantially lower latency.
By Sihan Ren, Gaozheng Li, Yuanshang Quan, Yiming Qin, Fuyi Yang, Chang Liu, Lan Xu, Minye Wu
The paper introduces Mawqif-XT, a new Arabic benchmark dataset comprising 996 manually annotated tweets from three public targets: Women Driving, E-Cars, and Trimester System. Each tweet is labeled for stance, sentiment, and sarcasm following the Mawqif annotation scheme, and the dataset is intended as a held‑out evaluation set to test cross‑target generalization. Baseline results are provided using Arabic and multilingual transformer models as well as zero‑shot large language models, enabling reproducible evaluation alongside the original Mawqif dataset.
By Rasha Albalawi, Nuha Albadi, Hamzah Luqman, Maram Kurdi, Saad Ezzini, Asma Yamani, Ahmed Ashraf
The paper introduces GAW-PO, a gradient‑aligned token reweighting technique for Direct Preference Optimization (DPO). It assigns weaker penalties to rejected tokens whose gradients align with preferred behavior, while maintaining stronger penalties for conflicting tokens. Across 11 benchmarks in mathematics, reasoning, coding, and question answering, GAW‑PO outperforms standard DPO and other baselines, and remains robust when the DPO regularization parameter is reduced.
By Andreea Dutulescu, Stefan Ruseti, Mihai Masala, Traian Rebedea, Mihai Dascalu