Policy optimisation, reward modelling and RLHF — how models are trained by feedback rather than by labels.
arXiv:2609.13457v1 Announce Type: new Abstract: Timeseries multimodal large language models (TS-MLLMs) have recently begun leveraging the reasoning capabilities of large language models (LLMs) for qu...
Zonal RL-RRT is a new path‑planning algorithm that partitions a map into zones using kd‑tree partitioning and employs Value Iteration as a high‑level decision maker. The method achieves a three‑fold improvement in time efficiency over basic sampling methods such as RRT and RRT* in forest‑like maps, and outperforms heuristic‑guided methods like BIT* and Informed RRT* by 1.5× in runtime while maintaining robust success rates across 2D to 6D environments. It also shows on average a 1.5× better performance than learning‑based methods such as NeuralRRT* and MPNetSMP, and has been validated in simulations of a UR10e arm manipulator in MuJoCo.
The paper introduces a label‑free test‑time training approach for GUI grounding, leveraging confidence patterns in coordinate tokens rather than full‑sequence confidence. It proposes Confidence‑Anchored Learning (CAL) to filter pseudo‑labels and assign distance‑based binary rewards, and extends this to Confidence‑Anchored Negative Learning (CANL) which optimizes solely on negative samples to avoid noisy positives. Experiments show that CANL‑7B achieves 92.1% on ScreenSpot‑V2 and 33.8% on ScreenSpot‑Pro, improving the base model by 8.9%.
The paper proposes Bypass Observation, a non‑intrusive layer‑wise readout architecture that attaches read‑only observation heads to selected Transformer layers without feeding their outputs back into the backbone. Three variants are explored: a shared language‑model head across layers, layer‑specific heads, and a layer‑ or step‑adaptive head. The authors provide a closed‑form overhead estimate (≈ V/(12d)) and discuss ways to reduce cost, while distinguishing bypass chain‑of‑thought from conventional chain‑of‑thought and outlining potential applications to looped and recurrent‑depth Transformers.
arXiv:2609.15297v1 Announce Type: new Abstract: In order to make Reinforcement Learning algorithms applicable in real world scenarios, safety must be ensured even under adverse operating conditions....
arXiv:2609.14896v1 Announce Type: cross Abstract: A notable byproduct of LLM alignment training is mode collapse: the progressive loss of output diversity that narrows a model's expressivity at infer...
arXiv:2609.14557v1 Announce Type: new Abstract: When performing parallel data transmission through a network using multiple paths, it is practically important to minimize the maximum transmission tim...
arXiv:2609.15987v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models (LLMs). We introduce Bellman Policy...
arXiv:2609.14327v1 Announce Type: new Abstract: Variance penalization is a principled approach to risk-sensitive reinforcement learning (RL) that explicitly trades expected return for policy stabilit...
arXiv:2609.15169v1 Announce Type: new Abstract: Driving vision-language-action (VLA) models increasingly reason before acting, but their intermediate reasoning is often weakly grounded in physical sc...
arXiv:2609.13219v1 Announce Type: cross Abstract: Neural correlates of spatial cognitive map are well documented, yet exactly how neural circuits perform spatial navigation in complex environments -...
The paper introduces Drift-Constrained Optimization (DCO), a framework that treats behavioral drift during fine‑tuning of instruction models as a bounded constraint rather than an uncontrolled side effect. By defining a drift budget, the authors reformulate fine‑tuning as a direction‑selection problem, showing that choosing different update directions can qualitatively change outcomes. Experiments on Qwen3 models demonstrate that carefully selected directions improve scientific reasoning and multilingual translation while preserving reasoning capabilities and general performance.
The paper introduces a Large Language Model–enhanced Multi-Agent Reinforcement Learning framework for optimizing electric vehicle charging, station profitability, and grid stability in public charging systems. By using an LLM to select interpretable features from IoT data and dynamically balance conflicting objectives, the approach unifies grid, EV, and station optimization in a single loop. Experiments show the method outperforms existing baselines, improving market efficiency and cutting training time by more than 70%.
arXiv:2604.07487v2 Announce Type: replace Abstract: Large language model agents rely on effective model context to obtain task-relevant information for decision-making. Many existing context engineer...
CodeTS introduces a verifiable framework for generating time series from natural language by translating textual temporal descriptions into executable code, which then produces the desired series. The approach constructs aligned Text‑Code‑TS triplets for supervised initialization and employs multi‑stage execution‑based rewards to ensure code validity and time‑series quality. Experiments on eight benchmarks show that CodeTS outperforms both LLM‑based and supervised generative baselines, offering a strong zero‑shot solution for Text‑to‑TS generation.
arXiv:2609.13243v1 Announce Type: cross Abstract: We present GzDRL, a novel single-process reinforcement learning (RL) framework for Gazebo that overcomes longstanding bottlenecks in scalable, reprod...
arXiv:2609.13916v1 Announce Type: new Abstract: We present North Small Translate, an open-weight, LLM-based machine translation (MT) model with instruction-following capabilities built on the same fo...
arXiv:2602.08194v2 Announce Type: replace-cross Abstract: Open-ended learning frames intelligence as emerging from continual interaction with an ever-expanding space of environments. While recent adv...
The paper introduces TTIQ, a test‑time reinforcement learning framework that improves vision‑language model (VLM) adaptation by explicitly measuring image‑question dependence. By teacher‑forcing responses on the original and ablated image‑question pairs, TTIQ derives token‑level likelihood changes to estimate how much each input contributes, then uses these signals to construct a reward that favors jointly grounded, confident responses. Experiments on eight VQA datasets and various VLM sizes show that TTIQ consistently outperforms prior consensus‑based methods and generalizes across model families and unseen datasets.
arXiv:2609.14648v1 Announce Type: new Abstract: Aligning multi-turn dialogue agents is usually framed as matching turn-level human preferences, yet direct optimization of long-term outcomes is often...