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 26 of 54 · The Research front page

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

Portable RGB demonstrations can train mobile manipulation policies and reduce base-velocity errors

Yutian Zhang et al. propose VOMMI, a framework that collects portable demonstrations from RGB video of body and hand cameras for mobile manipulation. Their R2-VO module refines offline trajectories using sparse geometric anchors, and the resulting policy, post-trained only on portable demos, achieves 18.2% lower base-velocity error than a policy trained on robot-collected demonstrations, with comparable end-effector accuracy.

·2 min
Generative media

Splitting diffusion distillation into two phase-specific half-sized experts beats a full-size student.

The authors propose Phase-wise Velocity Distillation (PVD), which splits the diffusion generation process into a structural phase and a refinement phase, each handled by a half-sized expert. Using two half-sized phase-specific experts outperforms a single full-size monolithic student under the same one-forward-pass compute budget. PVD achieves FID 1.48 on ImageNet 256x256 and competitive results on T2I benchmarks while reducing active parameters by about 50%.

·2 min
Language models

Separate short- and long-term matrix memories improve recurrent network performance

Regan et al. introduce the Continuous Memory Machine (CMM), a recurrent architecture with two matrix-valued memory states for short- and long-term storage, updated jointly by a shared Transformer. The CMM outperforms a suite of baselines across algorithmic, in-context learning, and maze-solving tasks, and shows better generalization than prior memory-augmented networks while preserving interpretable attention patterns.

·2 min
Agents

Step-by-step process delivery boosts LLM predictability and auditability

Schabert and Peters propose delivering Standard Operating Procedure steps one at a time via the Model Context Protocol, rather than embedding the full SOP in the system prompt. Across 15,475 trials on SOP-Bench, this step-level approach raises process adherence from 76–95% to 95–99% and reduces ungrounded correct answers from 2.1–4.5% to 0.2–0.3%, while lightweight executors gain 6.5 percentage points in grounded accuracy.

·3 min
Reinforcement learning

Empowerment Leads Agents to Central States, but Not Always Adaptability

Ji and colleagues give geometric and temporal interpretations of empowerment, an information-theoretic measure of how strongly an agent can control its future state. They show that maximizing potential empowerment directs agents toward central, broadly accessible states in tabular Markov decision processes, but they also construct settings where high empowerment provides little ability to adapt to new rewards.

·2 min
Agents

Spoken dialogue memory jointly tracks who said what and when

The authors present PERSIST, a persistent memory system for multi-session, multi-speaker spoken dialogue that retrieves past conversational events by jointly scoring semantic content, acoustic speaker identity, and temporal state. On their new SpokenTrace benchmark, it reaches 85.08% end-to-end task accuracy and improves all-support EM@3 from 49.01% with a strong semantic retriever to 82.10%. By reusing intermediate dialogue-backbone representations, PERSIST avoids re-encoding query audio and reduces retrieval latency from 578.42 ms to 7.03 ms.

·2 min
Agents

Personal LLM agents improve recommendations using cross-platform history

The authors formalize a new recommendation paradigm in which a personal LLM agent uses cross-platform user history to mediate a platform's candidate ranking before delivering the final top-K slate. They introduce MediateRec, a benchmark with proxy and real cross-platform environments, and propose Personal Attribution Mediation Optimization (PAMO), a training method that counterfactually masks cross-platform history to allocate advantage mass while preserving platform-relative value. Experiments show that PAMO consistently improves rescue-harm balance over outcome-only reinforcement learning, yet even strong proprietary LLMs introduce non-negligible harmful overrides.

·3 min
Code & software

Benchmark shows coding agents struggle to synthesize static-analysis checkers

The authors introduce CheckerBench, an executable benchmark of 300 real-world tasks for long-horizon static-analysis checker synthesis, and CheckerLab, a unified evaluation framework. Across 21 model-harness configurations, the mean Pass@1 is 32.30%, with the best configuration reaching 45.33%. The results indicate that reliable, reusable checker development remains difficult for current coding agents.

·2 min
Language models

Controlling gradient admission stabilizes LoRA fine-tuning of small language models

The authors propose GRADE, a data-centric framework that controls which data-induced gradients enter the LoRA subspace during SLM fine-tuning. Across three backbones and seven heterogeneous instruction datasets, GRADE consistently improves accuracy over standard LoRA and outperforms existing selection and stabilization methods, demonstrating that successful adaptation depends on managing gradient admission under constrained update capacity.

·3 min
Theory

Learned activations reshape spectral tuning in neural fields

The authors introduce learnable spectral activations (LSA), which replace fixed nonlinearities in implicit neural representations (INRs) with a layer-wise truncated Fourier series whose harmonic amplitudes are learned alongside network weights. This decouples feature selection from spectral shaping, concentrating target-signal energy in top eigenmodes of the neural tangent kernel and improving reconstruction across audio, image, and neural field tasks.

·3 min
Evaluation & benchmarks

Pass@k misses output diversity changes caused by model post-training

Rath and colleagues compared two post-training methods applied to the same Qwen2.5-1.5B-Instruct checkpoint on grade-school mathematics. Across three seeds, the methods moved token entropy, answer entropy and unique-answer counts in opposite directions, yet pass@8 and pass@32 did not consistently distinguish them. Comparisons with the original checkpoint also showed that GRPO’s pass@1 advantage over rejection-sampling fine-tuning reflected better capability retention, not a demonstrated capability gain.

·3 min
Reinforcement learning

Low-rank gradient sketches cut memory costs in language-model reinforcement learning

Zhang et al. present LoGRA, a reinforcement-learning post-training method that represents selected weight gradients as compact low-rank sketches instead of storing dense gradients and optimizer states. Across reasoning experiments, the authors report memory reductions of up to 45.7 percent without lower task performance, and sustained training of a 27B-parameter model under a hardware allocation where dense Adam could not fit.

·3 min
Robotics & embodied AI

Retrieval replaces generation for building simulation-ready articulated objects from a single photo

The authors present AffordCraft, a method that constructs a physically valid articulated simulation asset from a single RGB image and a task instruction by retrieving and adapting a matching entry from a library of pre-built assets, rather than estimating parts and joints from scratch. On 2,000 photographs across 31 categories, AffordCraft yields a valid asset for 85% of inputs, compared to at most 45% for five generative methods, while using 10–78× less GPU time per valid asset.

·3 min
Safety & security

Privately learning public data mixtures improves differentially private training

Chen, Afonja, Thudi, and Papernot present DP-MixMin, a privacy-preserving method that learns the best mixture of public datasets for pretraining before differentially private fine-tuning on sensitive data. On NIH ChestX-ray14, the learned mixture boosted macro AUC by up to 0.037 across privacy budgets, with a 22.8% relative gain on Cardiomegaly at epsilon=1; on ENRON, the learned mixture reduced test perplexity by 16% relative to baseline mixtures.

·3 min
Generative media

New estimator achieves optimal sample complexity for FID evaluation

The authors provide a comprehensive theoretical analysis of the sample complexity of estimating the Fréchet Inception Distance (FID), a standard metric for evaluating generative models. They introduce a new estimator, Relative Taylor Debiasing (RTD), which achieves the minimax optimal sample complexity of O(d/ε²) for d-dimensional image embeddings, proving it is impossible to improve upon. Their experiments on synthetic Gaussians and ImageNet confirm that RTD reduces estimation error, and a second-order variance-aware estimator (VALE₂) matches the accuracy of the prior state-of-the-art with only 10,000 samples instead of 50,000.

·3 min
Agents

Bidirectional protocol connects LLM agent workflows with inference engines

The authors propose HEAR, a bidirectional protocol that defines four semantic categories for information exchange between an agent harness (which manages workflow dependencies and context) and an inference engine (which handles request batching and KV-cache). Implementing two coordination policies—cache-aware runtime scheduling and role-specific inference configuration—the authors demonstrate on four benchmarks that combining harness intent with engine state improves throughput and latency without degrading task quality.

·3 min
Language models

Bayesian memory module adapts reasoning strategies during test time

The authors introduce the Bayesian Cheatsheet, a structured memory module for large language models that uses a Hierarchical Dirichlet Process Gaussian Mixture Model (HDP-GMM) to organize reasoning behaviors into an unbounded number of clusters. In online test-time training, the module retrieves relevant clusters for a query, updates cluster statistics after each sample, and can create new clusters for novel behaviors. On reasoning benchmarks like AIME'25, Omni-MATH, and PhysReason, Bayesian Cheatsheet outperforms existing memory modules, even in cold-start settings.

·3 min
Speech & audio

Choosing speech encoder layers on evaluation data inflates depression scores

Perez-Toro and colleagues examine a common evaluation shortcut in speech-based depression detection: selecting an encoder’s best-performing layer using the same cross-validation results reported as final performance. Their repeated nested evaluations show that this practice can raise apparent area under the ROC curve (AUC) by as much as 0.09, including substantial gains when the outcome labels contain no real signal.

·3 min
Theory

Single-pass backpropagation learns Pareto front at cost of standard training

The authors introduce MOSEL, a framework that learns a set of Pareto stationary solutions from a single forward-backward pass. It leverages bilevel optimization and Stackelberg game theory to decouple representation learning from preference alignment. Empirically, MOSEL recovers diverse Pareto frontiers in strongly conflicting tasks and converges closer to the utopia point in weakly conflicting settings, matching the efficiency of single-objective training.

·2 min
Evaluation & benchmarks

Multilingual benchmark exposes deepfake detectors’ reliance on synthetic audio cues

Segna, Tschesche and Rohrbach introduce a consent-based benchmark for testing audio-visual DeepFake detectors across languages, manipulation methods and demographic groups. Their experiments show substantial detection failures when manipulated video retains authentic audio, suggesting that some multimodal systems rely heavily on audio-video inconsistencies rather than robust visual evidence.

·2 min
Robotics & embodied AI

Caching frozen model states makes forward-only robot adaptation much faster

Kim and colleagues present VLA-ZO, a forward-only method for adapting vision-language-action models to deployment shifts using one target-environment demonstration. On LIBERO tasks with altered camera viewpoints, the method made zeroth-order optimization up to 32.54 times faster than a naive implementation, while raising average task success from 48.27 percent without adaptation to as high as 63.58 percent.

·2 min
Efficiency & systems

Centering activations enables calibration-free 4-bit diffusion transformer quantization

Jovanović, Salzmann and Javed identify why data-free Hadamard rotation performs poorly when quantizing diffusion transformers: adaptive layer normalization introduces a token-specific mean that rotation concentrates into a few range-dominating coordinates. Their CentriQ method removes this mean before quantization and restores it through a full-precision rank-1 computation, enabling calibration-free quantization of both weights and activations.

·3 min
Agents

LLM aggregators partly count restated evidence as new readings, measure shows

The authors develop a metric that converts an LLM aggregator's reported probability into the number of independent readings it implies, so a restated copy receives a weight between 0 (counts only sources) and 1 (counts every statement). Testing four LLMs across three testbeds—controlled logs with an exact Bayesian oracle, web documents with appended copies, and agent-written logs under four communication protocols—they find that a forwarded copy counts for 0.06 to 0.42 of a new reading. A one-paragraph declaration of what a copy contributes reduces copy weight to 0.08 or less on controlled logs, and a rule that has agents refer to readings instead of restating them cuts belief-implied early commitment from 11.2% to 1.1%.

·3 min
Safety & security

Activation steering shifts how language models decide to trust users

Lasnier and colleagues study trust from the assistant’s perspective: whether a model acts on claims whose accuracy or intent it cannot verify. Using 2,000 paired conversations, they identify activation-level interventions that monotonically increase or decrease trust-related behavior without changing model parameters, then test whether those interventions transfer to safety-relevant agent scenarios.

·3 min
Agents

Automated observer catches fail-plausible LLM errors but misses novel ones

The authors built an automated user-viewpoint observer targeting fail-plausible silent failures in a production LLM agent runtime. The two-layer pipeline—five deterministic signals distilled from past incidents escalate to an LLM judge that must cite verbatim evidence—achieves 6/6 detection on regression incidents with 0 false positives, but 0/4 recall on novel patterns. A pre-registered live deployment over 12 days fired zero verdicts, reported as undefined precision per the protocol.

·2 min
Efficiency & systems

Multi-precision LLM weight format keeps a single prefix-readable stream

The authors present StagQ, a multi-precision weight format for LLMs that stores weights in a single prefix-readable stream: a 2-bit group-wise affine base followed by configurable 1-bit refinement planes. A sparse side record holds out extreme weights via two guard stages. On Llama-3.1-8B, Phi-4, and OLMo-2-7B, StagQ-XXS leads the strongest multi-precision baseline on MMLU by 3.1 to 7.0 points at 2 bits, with faster batch-one matrix-vector products on an A100 GPU in most cases.

·3 min
Agents

MiniCorp: a simulator for studying AI-run companies and generating enterprise data

The authors present MiniCorp, a simulator that couples an internal world of persistent agent roles with an external world of customers, competitors, and market mechanisms. Using an e-commerce company as a demonstration, they show that agents can coordinate, adapt to market feedback, and sustain long-term strategies. The simulator generates continuous records of decisions and outcomes, and checkpointing enables replay under alternative decisions, providing counterfactual data unavailable from historical archives.

·3 min
Agents

Selective KV-cache reuse depends on request history in rolling agent workloads

The authors show that when a KV cache persists across multiple LLM calls in a rolling agent workload, the same prompt can produce different answers depending on the sequence of earlier requests. With a fixed 5% recomputation budget, document-aligned recomputation reduces answer variation across request orders from 69.0% (with CacheBlend's token top-k policy) to 26.1%. It also improves fidelity to full prefill by 34.5–52.5 percentage points while both policies achieve approximately 5.7× median time-to-first-token speedup.

·3 min
Vision

Vision-language models detect object rotation but misjudge its magnitude

Wang and colleagues evaluated 12 vision-language models on eight tasks requiring them to detect, estimate and reason about object rotations across views. Models recognized whether rotation occurred, but magnitude estimates remained near chance and exact-angle answers collapsed onto 0°, 90° and 180°. The authors show that some coarse rotation information is present inside model representations but is not effectively used during answer generation.

·2 min