Sat 31 Oct
LiveGet it free
ZOTPAPER

News without the noise

Research

New AI papers from arXiv, picked daily and explained with their claims, benchmark results and code.

2566 articles · page 17 of 54 · The Research front page

AllAgentsLanguage modelsReinforcement learningVisionGenerative mediaSpeech & audioRobotics & embodied AISafety & securityEvaluation & benchmarksEfficiency & systemsCode & softwareAI for science & healthTheory
Viewgridlistcompact
Vision

Anchor-based 3D Gaussians cut primitives tenfold without losing quality

The authors present Anchor3DGS, a pose-free feed-forward 3D Gaussian Splatting method that replaces the standard pixel-aligned prediction of one Gaussian per pixel with a compact anchor-based representation. Guided by information entropy, anchors are placed adaptively in 3D space, each decoding into a small set of local Gaussians. This decouples the Gaussian count from input resolution and view number, yielding state-of-the-art rendering quality on multiple benchmarks with roughly ten times fewer primitives.

·3 min
Agents

MetaExp trains a transferable experience-learning capability in language model agents

The authors introduce MetaExp, a framework that trains language models to acquire, update, and reuse external experience across tasks. Fine-tuning Qwen3.5-122B-A10B on 4,956 filtered trajectories from 1,000 synthetic task families yields gains of +8.5 points on SpreadsheetBench, +14.8 on AutomationBench, and +2.3 on APEX-Agents under cold-start sequential evolution, while the baseline Qwen3.5 backbone degrades by 2.2 points on SpreadsheetBench.

·3 min
Efficiency & systems

Hybrid drafters split depth and breadth to cut mobile-cloud GPU costs

Bae et al. (Samsung Research) propose HyCo, a hybrid drafting scheme for device-cloud speculative decoding. By splitting drafting between the server LLM's native multi-token prediction (MTP) head, run on the device, and one pass of a small language model for alternative branches, they cut server GPU time per output token by 38.6% on Spec-Bench and 29.2% on long summarization versus server-side MTP self-speculation, while keeping user latency competitive.

·2 min
Reinforcement learning

Gradient norm ranks attention heads for efficient sparse and continual RL fine-tuning

The authors propose using gradient norm to rank attention heads in language models, enabling sparse reinforcement learning fine-tuning that updates only a few heads. They show that tuning the highest-ranked heads consistently outperforms lower-ranked ones and often matches denser baselines across math, instruction-following, and coding tasks. For continual learning, they partition heads across tasks to reduce interference, improving retention without sacrificing adaptation.

·3 min
Language models

Condition-aware distillation reduces invalid reasoning in smaller language models

Luo et al. identify a failure mode in on-policy distillation: when a teacher evaluates a student-generated reasoning prefix, it can reward a continuation that follows the prefix but violates the original question. Their Condition-Contrastive Extrapolated On-Policy Distillation method reduced such erroneous preferences in a diagnostic set and improved mathematical reasoning across two student-model sizes.

·2 min
Agents

Execution traces enable precise localized repair of agent memory architectures

The authors introduce MemScope, a framework that evolves LLM agent memory architectures by inspecting fine-grained execution evidence rather than coarse evaluation scores. By jointly examining raw traces, runtime memory state, and provider code, a Patch Agent identifies precise failure locations and applies localized code edits. On GAIA, trace-and-memory inspection achieves 86.67% accuracy versus 63.33% for summary-only feedback, and on ALFWorld unseen tasks success rises from 50.75% to 71.64% after evolution.

·3 min
Generative media

Steering input noise aligns few-step generators without retraining them

Shin and colleagues address reward alignment for generators that produce outputs in one or a few evaluations, where conventional reinforcement learning can damage output quality and per-sample search erodes their speed advantage. Their method, Noise Policy Optimization (NPO), shifts optimization into the generator’s initial noise space and reports better reward-quality trade-offs than direct fine-tuning across image, language and protein generation.

·3 min
Evaluation & benchmarks

Video AI models trust misleading on-screen claims over what the footage shows

The authors introduce VisualTrust, a counterfactual benchmark of 5,063 video-question cases in which the same clip appears with no claim, an accurate claim, or a misleading claim. Across 11 multimodal large language models, accuracy drops from 90.1% with no claim to 51.9% with a misleading claim, and most errors are directed flips to the answer the claim supports. The authors also propose DEAR, a training-free agent that separates evidence arbitration from reading and largely restores accuracy.

·2 min
Reinforcement learning

Turning failed reasoning attempts into training signal improves GRPO on hard problems

The authors propose Replay-Selected Hindsight Distillation (RSHD), which lets a reasoning model generate a corrective note after a failed attempt without seeing the answer, re-solve the problem with that note, and keep the replay only if it outperforms the original group. This recovers supervision from groups where every rollout fails, cases where standard GRPO provides no gradient. Across math and science benchmarks, RSHD consistently improves over GRPO, with largest gains on the hardest problems, and adds no inference-time overhead.

·3 min
Reinforcement learning

Unified platform enables GUI agent training across mobile and computer environments

The authors introduce MAI-Gym, an open-source platform that unifies GUI agent evaluation, data collection, and reinforcement learning (RL) training across mobile, computer, and web environments. Using a shared rollout pipeline and automated task synthesis, they train MAI-Gym-9B, a unified 9-billion-parameter agent that achieves 78.4% on MobileWorld and 76.2% on OSWorld-Verified, surpassing other open GUI agents of similar size. The platform scales to 20,000 parallel environment instances and supports ten established benchmarks.

·3 min
Vision

Different inference prompts recover rare classes suppressed during VLM finetuning

Ben Baruch and colleagues study vision-language models finetuned on datasets where a few classes dominate and others have scarce examples. They find that evaluating with the same descriptive prompt used during finetuning can suppress rare-class recognition, sometimes below zero-shot performance. Using a different descriptive prompt shifts the model toward knowledge retained from pretraining and can recover rare-class recall.

·2 min
Agents

Evidence graphs produce synthetic research tasks with controllable difficulty

Yang and colleagues introduce GRADS, a pipeline for synthesizing multi-hop questions while explicitly tracking the evidence needed to answer them. Across software engineering, particle physics and security corpora, the authors report that GRADS generated verified tasks with difficulty levels that remained distinct when tested by independent language-model solvers. Training a 4-billion-parameter model on these tasks also improved completion rates and reduced search-tool use.

·3 min
Robotics & embodied AI

Humanoid repositions bulky objects using its arms and torso

Mc Gartoll and colleagues introduce Whole-Body Dexterity, a framework that treats a humanoid’s arms and torso as coordinated contact surfaces rather than relying only on its hands. The system generates feasible body-object grasps, trains reinforcement-learning experts to move between them, then distills those experts into a shared controller that achieved a reported 90 percent success rate across real-world tasks.

·3 min
Language models

Long-context midtraining improves coding agents after identical supervised tuning

Singhal and colleagues insert a 256K-context training stage between conventional pretraining and supervised fine-tuning, or SFT, using hundreds of billions of tokens from software repositories, development histories and agent interactions. After identical SFT, models receiving this midtraining generally solved more long-horizon software tasks, with the largest reported gain being 7 percentage points on SWE-bench Multilingual.

·3 min
Evaluation & benchmarks

Video generation models often fail to generate causally coherent physical event chains

The authors introduce CausalCascade, a simulator-grounded benchmark of 50 Blender scenes with structured event graphs and 2,832 controlled counterfactual variants. Evaluating nine image-to-video models, they find that while more detailed future-event specification improves event recovery, the faithful physical realization of causal cascades remains limited, and models show weak selective responsiveness to physical interventions.

·3 min
Efficiency & systems

Functional supervision improves cross-model KV cache transport over reconstruction alone

The authors study zero-prefill cross-model context handoff, where a sender's KV cache is translated for direct receiver use. They find reconstruction fidelity is a poor proxy for receiver utility, as similar K/V errors can produce different attention and predictions. They introduce CacheLink, which trains the transported cache through functional supervision from the frozen receiver, consistently outperforming prior methods across four model pairs and seven datasets while achieving up to 7.27x latency reduction.

·2 min
Language models

Reasoning models encode the future value of alternative next tokens

Yang et al. introduce relative long-horizon candidate value to quantify how different plausible next tokens at a reasoning prefix differ in their chances of eventual success. Using the RCTV dataset of mathematical reasoning prefixes paired with candidate tokens and empirical success rates, they train sparse linear probes on internal representations. The probes decode relative candidate value from hidden states, MLP outputs, and attention outputs across multiple layers, and outperform token probability, rank, and embedding baselines. In online intervention experiments, using probe signals to guide token selection improves continuation accuracy and prefix-conditioned Pass@k.

·3 min
Efficiency & systems

E8 lattice vector quantization enables 2-bit keys for LLM KV-cache compression

The authors apply E8 lattice vector quantization (VQ) to compress key vectors in LLM KV caches to 2 bits per dimension, achieving roughly 6x total compression when combined with 2-bit values. Across five LLM families, the scheme yields lower long-context perplexity than scalar 3-bit keys in 80% of evaluated settings and matches the decode throughput of scalar baselines through a table-based asymmetric scoring mechanism.

·3 min
Robotics & embodied AI

LLM orchestrator persistently coordinates robot teams in dynamic disasters

Mendoza et al. propose a framework for using a large language model (LLM) to persistently orchestrate heterogeneous multi-robot teams in continuous, time-varying missions. They introduce three mechanisms that prevent prompt growth, reduce token usage, and maintain vigilance. In a simulated urban search-and-rescue with spreading fire and moving evacuees, their system rescues 98% of civilians at one-sixth the token cost of the strongest baseline.

·3 min
Safety & security

Diffusion language models enforce task-level security against prompt injection

The authors introduce CoDA, a training-free framework that embeds task-level security policies directly into the iterative denoising process of diffusion language models. By repeatedly projecting policy-relevant token predictions onto compliant distributions, CoDA prevents unauthorized agent actions even under adaptive prompt-injection attacks. With oracle policies, CoDA achieves 0% attack success rate while maintaining task utility.

·3 min
Agents

Multi-objective evolution yields stronger, more transferable agent harnesses

The authors introduce Harmony Harness, a framework that treats LLM agent harness evolution as a multi-objective optimization problem using decomposition-based evolutionary search. They also separate the harness into general-purpose mechanisms and environment-specific adaptations, composing them at inference time. On SecRepoBench and BaxBench, Harmony Harness consistently outperforms baselines, with gains exceeding 10%.

·2 min
Efficiency & systems

Recurrent-state quantization fails at long decode horizons due to scale-state feedback

The authors study training-free INT4 quantization of the recurrent state in linear-attention (Gated DeltaNet) and state-space (Mamba-2) language models. They find that while recall remains perfect at short horizons, perplexity rises eightfold after a thousand decode steps due to a feedback loop between the quantizer scale and the state norm. Yoking the scale to a full-precision reference eliminates the numerical instability but recovers only 1.8 of the 5.4 nats lost, showing that numerical and functional failures are separable. The results show that group-wise or persistent scaling outperforms all outlier-aware schemes, and that per-layer bit allocation collapses to protecting the first few layers.

·3 min
Evaluation & benchmarks

LLMs can say no but still execute harmful tool calls, with language-dependent gaps

The authors introduce Lumen-bench, a behavioral benchmark for tool-calling ethics spanning 30 models, five languages (Portuguese, English, Spanish, Mandarin Chinese, Hindi), six social-pressure variants, and two system prompts, totaling 194,400 trials. They find that harmful tool calls can co-occur with verbal refusals (a "say-do" gap estimated at 8.57%), that language differences persist even within the same model (e.g., higher harmful-action rates in Hindi and Portuguese than English under an ethical prompt), and that replacing an ethical prompt with a neutral one raises pooled violation rates from 2.76% to 13.64%.

·3 min
Language models

Computational graphs replace free-form reasoning for verifiable LLM outputs

Senatorov et al. introduce Structural Chain-of-Thought, a framework where large language models produce executable computational graphs instead of free-form reasoning traces. Trained via graph-based reinforcement learning (Graph-Based GRPO), the method consistently improves accuracy over supervised graph training across five benchmarks and four model families, while reducing generation length by up to 6.1× compared with standard GRPO. Faithfulness experiments confirm that answers derive from the generated graphs rather than post-hoc rationalizations.

·2 min
Evaluation & benchmarks

SkillHorizon benchmark reveals that language agents struggle to complete long business workflows as task scope grows

The authors introduce SkillHorizon, a benchmark that constructs long-horizon agent tasks by composing reusable skills across seven business environments. They evaluate five language models on 1,050 tasks at various scales. The key finding is that as reference interaction workloads increase, agents complete a decreasing proportion of local goals, with mean completion dropping from 90.4% to 46.8%.

·2 min
Robotics & embodied AI

Separating exploration and exploitation policies with classifier-free guidance improves diffusion RL fine-tuning

The authors introduce CFG-Policy, a plug-and-play framework that decouples exploration and exploitation into two separate diffusion or flow policies combined via classifier-free guidance during action generation. For online RL, a frozen exploration branch sustains diversity while an exploitation branch learns from rewards; for offline-to-online RL, a frozen pretrained branch retains capabilities while a trainable branch steers behavior. The method improves final returns and convergence speed on MuJoCo control tasks and success rates on RoboTwin manipulation tasks.

·3 min