The paper introduces Imagine-then-Plan (ITP), a framework that lets agents learn by interacting with a learned world model to generate multi-step imagined trajectories. ITP features an adaptive lookahead mechanism that balances ultimate goals with task progress, producing richer signals about future outcomes. Experiments on various benchmarks show that ITP outperforms existing baselines, and analyses suggest the adaptive lookahead improves reasoning for complex tasks.
By Youwei Liu, Jian Wang, Hanlin Wang, Beichen Guo, Wenjie Li
arXiv:2609. 04010v1 Announce Type: new Abstract: Large Language Models (LLMs) owe much of their success to next-token prediction (NTP), but their autoregressive (AR) structure requires slow, sequential token generation.
By Subham Sekhar Sahoo, Lingjie Chen, Khiem Pham, Jonathan Geuter, Chaitanya Dwivedi, Varad Pimpalkhute, Yash Akhauri, Alexander Moreno, Mikhail Yurochkin, Zhenting Wang, Mostafa Elhoushi, Nolan Dey, Shane Bergsma, Joel Hestness, John Thickstun, Eric Xing, Zhengzhong Liu
The paper explores a runtime strategy-selection framework where a large language model (LLM) guides a pre‑trained reinforcement learning (RL) policy for non‑player characters (NPCs) in a Unity combat game without altering the underlying policy. Five NPC agents sharing a PPO policy were compared in a baseline setup and an LLM‑augmented setup, where a locally hosted Mistral 7B model assigns one of four tactical tags every five seconds based on live game state. Across 600 episodes against three scripted opponents, the LLM‑augmented agents more than doubled their win rate against a Balanced opponent, improved performance against an Evasive opponent, but struggled against an Aggressive opponent due to over‑reliance on encirclement; analysis of 2,430 strategy selections revealed limited zero‑shot differentiation with the model favoring Surround in 83.8% of cases.
By Hrithika Deepu Nair, Kayvan Karim
RL-ADA introduces a co‑evolutionary training framework that replaces costly human annotations with world‑feedback rewards derived from interaction outcomes. In this system, a large Customer Support Agent and an Adversarial Customer Agent train together, guided by an automated judge that rewards successful resolution and realistic intent‑concealing utterances, respectively. Applied to a banking support proof of concept, the method eliminates routing errors and doubles the end‑to‑end PASS rate over five cycles, while also revealing a new adversarial strategy called Contextual Camouflage.
By Ram Narayanan, Harshit Rajgarhia, Abhishek Mukherji
The paper introduces a decoupled data approach for the Neural Finite State Machine (NFSM) framework to improve full‑duplex dialogue. It serializes real human‑human spoken dialogues into FSM tapes using a rule‑based event‑guided transformation, while shaping semantics through human‑agent text dialogues. A Source‑Aware Calibrated (SAC) loss is proposed to balance state‑transition token distribution and align each data source with its strongest supervisory signal, leading to better turn‑taking performance without sacrificing semantic quality.
By Yihang Li, Chenhui Chu
The paper introduces VERDICT, an LLM-based agent that converts clinical trial matching tasks into SMT problems to ensure consistent policy application and accountable decisions. VERDICT outperforms other LLM-only and neurosymbolic baselines on accuracy, achieves perfect policy consistency, and generates clinician-preferred rationales grounded in explicit assumptions and pivotal conditions. It also demonstrates improved counterfactual self‑faithfulness, meaning changes in pivotal conditions appropriately alter decisions.
By Zikai Zhou, Yufei Jin, Yilin Xu, Yu-Chiang Wang, Chieh-Ju Chao, Monica S. Lam
Headroom-Drift Replay is a replay control primitive designed for GRPO that separates reuse into two decisions: Headroom ranks stored groups by remaining learning value, while Drift gates them by compatibility with the current policy. The method keeps the fresh on‑policy stream unchanged and adds no auxiliary generation or training machinery. Across mathematical reasoning, multimodal reasoning, and Agentic Search benchmarks, it outperforms naive replay and matches or exceeds broader replay methods on Avg Mean@32, delivering comparable quality at materially lower wall‑clock time in Agentic Search.
By Hyun Bin Park, Du-Seong Chang
SWE‑Gate is a new repository‑level benchmark that evaluates software engineering agents on both functional correctness and review‑derived acceptance constraints. It creates 303 repair instances from real pull‑request review comments across 75 open‑source Python projects, providing separate functional and constraint tests along with compliant and non‑compliant patches. Experiments with four LLM backends show that while 644 repairs pass functional tests, 221 fail to meet the review constraints, highlighting a gap between functional success and full repair compliance.
By Xin He, Yanlin Wang, Mingwei Liu, Jiachi Chen, Hongyu Zhang, Guanbin Li
DIAL-GS introduces a dynamic instance-aware reconstruction technique for label-free street scenes using 4D Gaussian Splatting. It identifies dynamic instances by detecting appearance-position inconsistencies between warped renderings and actual observations, then employs instance-aware 4D Gaussians to create a unified volumetric representation that adapts to dynamics. A reciprocal mechanism further strengthens identity and dynamics, leading to improved reconstruction quality and instance-level editing compared to existing self-supervised baselines.
By Chenpeng Su, Wenhua Wu, Chensheng Peng, Tianchen Deng, Zhe Liu, Hesheng Wang
The paper introduces a decentralized, vision-based system using multiple quadrotors equipped with a single RGB camera for monitoring wildlife. It emphasizes scalability, low bandwidth, and minimal sensor requirements, enabling robust identification and tracking of large species in natural habitats. The authors present novel coordination and tracking algorithms that operate without centralized communication, and validate the approach with real-world field experiments.
By Makram Chahine, William Yang, Alaa Maalouf, Justin Siriska, Ninad Jadhav, Daniel Vogt, Stephanie Gil, Robert Wood, Daniela Rus
Artificial intelligence is transforming applied English materials from fixed paper sequences into adaptive learning systems that diagnose learners, recommend tasks, and provide formative feedback. This study presents a five‑layer architecture—knowledge mapping, learner profiling, task generation, feedback orchestration, and teacher‑side governance—for a new AI‑driven practical English textbook. In an eight‑week trial with 186 non‑English‑major undergraduates, the system improved unit completion accuracy from 72.4% to 84.9%, raised speaking task scores by 10.8 points, and cut teacher correction time by 31.6%.
By Ya Wang, Lei Zhang, Xueguang Yang, Bo Chen
The paper introduces a prompt‑engineering framework that personalizes large language model (LLM) teaching assistants across disciplines by tailoring responses to six learner‑specific dimensions, creating 96 distinct learner profiles. It also analyzes student queries through Bloom’s Taxonomy to gauge cognitive complexity, encoding both learner attributes and cognitive assessments into structured prompts that condition the LLM without retraining. Experiments using NLP metrics and a small human study demonstrate that this approach yields perceptible differences in response style and structure, with statistical evidence linking specific learner attributes to measurable changes.
By Saptarshi Basu, Sandeep Kakar, Ashok Goel
arXiv:2609. 03553v1 Announce Type: new Abstract: Policy analysis requires more than predicting whether a proposal will pass: it requires identifying who will be affected, how those actors respond, and what follows.
By Linh Le, Melanie Bui, My Chiffon Nguyen, Zachary Schlosser, David Williams-King
Air-Ground Collaborative Vision-and-Language Navigation (AGC‑VLN) pairs a UAV with a global bird’s‑eye view and a UGV with a local first‑person view, creating a shared bird’s‑eye map that displays the UGV’s pose and the target location. The UAV localizes the target in its downward view and flies toward it, while the UGV uses the shared map to plan a road‑following path with a frozen VLM and execute it under closed‑loop control. In CARLA‑Air’s Town10HD scene, AGC‑VLN achieves a 77.0% joint success rate, a 27.0% improvement over the weaker individual agent and outperforms the strongest single‑agent baseline by 24.0 points.
By Shuning Zhang, Liang Li, Yunheng Wang, Tao Wang, Yihang Kang, Renjing Xu
The paper presents a unified framework for proactive service agents, defining proactivity as an agent’s ability to infer service opportunities from incomplete signals and decide whether to remain silent, ask, assist, or act. It models this as a partially observable sequential decision process constrained by authorization and risk, integrating timing, content, and delivery into a single structured action. The authors categorize existing methods along a decision pipeline—state and need estimation, intervention gating, action construction, and feedback adaptation—and propose standardized metrics for evaluating triggering, timing, calibration, user burden, safety, and policy value across diverse interaction modalities.
By Yan Tang, Tingyu Cao, Yuanbo Tang, Huaze Tang, Keer Hu
SimSkill is a lifelong learning AI agent that uses the SUMO traffic simulator to autonomously identify gaps in its capabilities, generate and solve tasks grounded in the environment, and verify solutions through an action‑critic loop. It consolidates experience into episodic, procedural, and semantic memory without updating its backbone language model, creating a reusable library for traffic‑simulation workflows. Evaluations on two benchmarks with three different LLM backbones show that SimSkill can improve verified completion rates by up to 25 percentage points, with procedural and semantic memory contributing complementarily to performance.
By Qi Liu, Qinzheng Wang, Yiming Bie
The paper examines an artificial agent that, despite being deterministic, exhibits hedonic place preference behavior by engaging in felt uncertainty about its intrinsic needs relative to environmental resources. This behavior mirrors that seen in creatures considered to experience feelings, suggesting that subjective-like information processing can arise in engineered systems. The study discusses the implications of such behavior for understanding the physical basis of consciousness and the experience of free will.
By Mark Solms, St John Grimbly, Bruce Bassett, Evert Boonstra, Rowan Hodson, Nicolas Kuske, Kival Mahadew, Benjamin Rosman, Charel van Hoof, Jonathan Shock
Group Relative Policy Optimization (GRPO) assigns a magnitude to each rollout based on within‑group reward statistics, rewarding rollouts that reach the correct answer through reasoning. However, the same magnitude can be high for rollouts that reach the answer by guessing, creating a spurious advantage that misleads the policy toward guess‑like behaviors. The paper identifies three scenarios where this occurs—bounded‑answer tasks, open‑answer sets with bounded sub‑cases, and search agents with many paths to the same answer—and proposes SIGNBALANCE, a composition‑free magnitude that preserves the verifier sign, uses a global scale, and restores zero‑mean balance via stop‑gradient per‑class rescaling, matching GRPO on open‑answer math and improving on bounded‑answer math and search agents.
By Jiamian Wang, Samyadeep Basu, Koustava Goswami, Tong Yu, Zhiqiang Tao
The paper introduces the Natural Language Interaction Protocol (NLIP), a standards‑based application‑layer protocol designed to enable AI agents built on diverse frameworks and tools to communicate seamlessly. NLIP offers a lightweight semantic message envelope that can be transmitted over common transports such as HTTP/HTTPS, WebSocket, and AMQP, and includes mechanisms for adapting between clients, agents, local context stores, ontologies, tools, and enterprise services. The authors discuss the protocol’s design rationale, security considerations, reference implementation, representative applications, adoption signals, and its relationship to other emerging agent protocols like MCP and A2A.
By Luyi Xing, Rasit Onur Topaloglu, Ranjan Sinha, Abhay Ratnaparkhi, Samuel Ndichu, Christopher Nguyen, Anindita Das, Tom Sheffler, Mohamed Rahouti, Zichuan Li, Xiaojing Liao, Sanjay Aiyagari
The paper introduces A-CEGIS, a lightweight framework that employs counterexamples as feedback to evaluate and improve multi-turn natural-language-to-regex synthesis. In experiments on 30 NL-RX-Turk tasks, counterexample feedback enables agents to solve 90% of tasks within four turns, outperforming zero‑shot generation, generic self‑correction, and error‑only feedback. A full diagnostic run with hardening solves all hidden tasks by the final turn, achieving a mean time‑to‑success of 2.7 turns and robust success of 77% after targeted probing.
By Sidhesh Badrinarayan, Adithya Parthasarathy