PyTorchKR Papers

최신 논문

총 523,801편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2611.04116v1cs.LG

Decomposing Rank Reversals under Temporal Nonstationarity: Alignment, Aggressiveness, and Instability

Under temporal nonstationarity, model performance rankings reverse across training horizons due to heterogeneous sensitivities to memory. While existing literature addresses adaptive lookbacks, effective model complexity, and empirical forecast evaluation in isolation, no unified framework explains the structural mechanics driving these rank reversals. We bridge this gap by deriving a model-agnostic decomposition of out-of-sample $R^2$ into target alignment, miscalibrated aggressiveness, and predictive instability. Within this framework, we mathematically characterize complexity phase boundaries, establish closed-form optimal-memory equations for ridge regression under local drift, and extend these guarantees to complex model architectures via surrogate bounds. We validate these theoretical dynamics by benchmarking 14 linear, tree-based, and neural architectures across large-scale financial panels. In doing so, our framework demonstrates that empirical model rank reversals are driven by predictable surges in miscalibrated aggressiveness, providing a rigorous mechanistic diagnosis of generalization under temporal shift.

arXiv:2611.04113v1cs.AI

The Evolution Strategies Paradox: Better Models from Worse Perturbations

How is it possible for perturbation-based methods like Evolution Strategies (ES) to optimize billion-parameter large language models (LLMs)? Classical search methods are highly dependent on finding good candidates within generated populations. However, when using ES to fine-tune LLMs, a phenomenon can occur where all offspring models consistently perform worse than the parent model, yet during training, the parent model consistently improves. This paradox occurs across models, tasks, and even modalities. This paper characterizes the paradox, then examines how the local reward landscape and properties of the ES estimator contribute to it: in high-dimensional spaces, exploration leads to worse performance, yet still highlights good directions; aggregating those directions then allows for making local improvements. The result can be seen as a new optimization paradigm where a principled interaction of noise and curvature produces the paradox as a byproduct of effective optimization. These findings provide guidance for taking advantage of perturbation-based methods in LLM fine-tuning and in high-dimensional optimization in general.

arXiv:2611.04112v1cs.CV

Co-Occurrence Priors Override Visual Spatial Evidence: Understanding False Alarms in Vision-Language Models for Construction Struck-By Hazard Detection

Vision-language models (VLMs) increasingly support construction hazard monitoring, yet their frequent false alarms remain poorly understood. This study investigates whether false struck-by hazard warnings arise from failures in object recognition, spatial relation encoding, or hazard decisions dominated by learned co-occurrence priors. We evaluated Qwen2.5-VL-7B and InternVL3.5-8B using 355 counterfactual worker-excavator scene pairs with controlled spatial proximity and worker-removal variants, combining behavioral analysis, internal representation probing, causal interventions, and decision correction. Both models recognized the entities and encoded their spatial relations (out-of-fold AUROC: 0.82 and 0.91), assigning higher hazard scores to near than far scenes in 79% and 73% of pairs. Nevertheless, 74% and 100% of far scenes were falsely classified as hazardous. Removing the worker reduced Qwen's hazard rate to 11%, while worker presence exerted 2.4- and 3.2-times greater influence on hazard scores than spatial proximity in the respective models. Attention-head interventions, neuron-level interventions, and activation steering failed to eliminate this co-occurrence bias. However, conditioning hazard decisions on the model's own spatial-relation representation improved Qwen's specificity from 0.36 to 0.58 at matched sensitivity, outperforming threshold recalibration. These findings reveal that VLM struck-by false alarms stem primarily from decision-making failures rather than deficiencies in object recognition or spatial encoding, suggesting that future efforts should prioritize conditioning hazard decisions on encoded spatial evidence rather than further improving visual recognition alone.

arXiv:2611.04111v1cs.CV

Generalizable AI-Generated Image Detection with Embedding-Space Explanations

AI-generated images have reached unprecedented realism, while instruction-driven AI editors re-synthesize images for local changes, leaving no splice boundary to localize. Most detectors are black boxes with unverifiable verdicts and generalize poorly because adapting a backbone on the detection distribution memorizes generator-specific fingerprints that quickly become obsolete. Specialist detectors built on pure-vision representations, particularly DINOv3, generalize well across generators, but explainable methods forgo this advantage: jointly training one multimodal model for classification and language generation subjects its visual representation to the language objective. We therefore require that explanation training cannot alter the verdict. We introduce FOX, which first builds and locks a specialist DINOv3 detector, then uses its representation to guide a lightweight VLM toward forensic explanation. A joint-embedding predictor maps locked visual evidence into a frozen text encoder's semantic space, producing guidance that the VLM verbalizes on demand. FOX achieves 97.4 mean AUC across 14 unseen generators, 69.8 BERTScore on XPlainVerse, and 56.1 on LOKI.

arXiv:2611.04107v1cs.CL

Say Less, Mean Better: Training-Free Verbosity Control for Language Models

Large language models (LLMs) often generate responses longer than necessary, increasing inference latency and computational cost. While prompting can encourage conciseness, controlling output length without compromising response completeness remains challenging. Aggressive early stopping can reduce generation costs but risks terminating responses prematurely. We introduce MinimalLM, a training-free framework for reliable verbosity control that combines uncertainty-aware, length-dependent stop-token adjustment with structural completeness guardrails. MinimalLM separates the decision of when to stop from whether stopping is structurally appropriate, coordinating response-level instructions, decoding-time intervention, and completion checks without modifying model parameters or requiring auxiliary models. We evaluate MinimalLM across three open-source LLMs and 330 questions, producing 5,940 responses across baseline and ablation configurations. MinimalLM reduces generated tokens by 47.6%, accelerates individual responses by 19-68%, and eliminates structurally truncated outputs in the primary evaluation. Despite substantial token savings, it achieves an average quality score of 8.04/10, compared with 8.42/10 for unmodified models and 6.46/10 for length-matched truncation. Evaluations across diverse subject domains (such as healthcare, programming, law, and science) demonstrate robust efficiency gains, while ablations establish the complementary roles of decoding control and completeness constraints. Our findings highlight that efficient generation requires more than producing fewer tokens: it requires controlling when generation ends while preserving response integrity. MinimalLM offers a lightweight, training-free approach toward this goal.

arXiv:2611.04106v1cs.LG

Structure, Dimensionality, and Distribution Shape - What Makes Pretraining Data Transferable, and Why the Input Pipeline Decides Which

What makes a pretraining corpus transferable? We describe a corpus by three properties: its structure, its dimensionality, and the distribution shape of each unit's values. We propose the Transfer-Invariance Principle: the same property can be irrelevant to transfer under one input transform and decisive under another, so the model's input transform, not the data modality, decides which properties can matter. To test it we build TransferForge, a generator that places a synthetic population at any prescribed structure, dimensionality, and distribution shape. KernelSynth, the synthetic prior of Chronos, cannot, because it draws independent series. When the model standardizes each input window (PatchTST on UCR), the distribution shape is inert and the structural setting governs. Synthetic pretraining then matches a real corpus of the same size and exceeds it once scaled. When the model encodes calcium fluorescence signals without normalization (POYO Plus), the shape becomes a governing constraint, and a shape-matched synthetic corpus replaces real pretraining with zero real neural activity. Flipping only the input normalization, with the model and data fixed, changes the effect of a shape manipulation by a factor of 5.3, so the pipeline, not the modality, sets the switch. The principle yields a recipe that chooses what to synthesize before any training. Pretraining data then becomes something to manufacture for the model, most valuably where real data is scarce, expensive, or private.

arXiv:2611.04104v1cs.LG

ModelBlaster: Closing the Performance Gap through Learning-Augmented Agents Across Models, Kernels and Hardware Backends

Optimizing accelerator code across hardware architectures is difficult. ML compilers search predefined spaces of schedules and micro-kernels, while LLM-guided optimization can explore alternative algorithms, operator decompositions, and hardware-specific techniques. However, existing LLM agents typically target a single programming interface, toolchain, or accelerator family, focus on kernel-level optimization without model-level transforms, and offer limited reuse of experience across hardware targets. We present ModelBlaster, a Memory-Augmented In-Context Reinforcement Learning (MAIC-RL) framework for optimizing data-parallel kernels and whole models across heterogeneous SIMD and SIMT accelerators. Its mutable policy is a persistent knowledge base updated at inference time by Actor-Critic Textual Gradient Descent (AC-TGD). The knowledge base maps performance states to actions, outcomes, and constraints, enabling experience reuse across kernels, models, and hardware targets. A model-level optimizer uses kernel-level knowledge to inform whole-model decisions. The same optimization loop and knowledge schema support CUDA on NVIDIA GPUs, OpenCL on Qualcomm GPUs, and RVV intrinsics on RISC-V vector cores, with only lowering and measurement layers specific to each target. Compared with PyTorch, ModelBlaster achieves geometric-mean speedups of 1.85x, 5.58x, and 5.18x on KernelBench Levels 1, 2, and 3, and 1.69x on SOL-ExecBench Levels 1 and 2 on a B200 GPU. On Qualcomm Adreno GPUs, it achieves 5.8x over the vendor SDK and 1.7x over Apache TVM; on RISC-V vector cores, it achieves 2.7x over ExecuTorch with XNNPACK across 70 KernelBench problems.

arXiv:2611.04102v1cs.CV

Is This 3D Asset Usable? Beyond the Beauty Render for Asset Curation

Large 3D repositories make it easy to collect assets, but difficult to determine whether they can be reused reliably across scene assembly, editing, rendering, game production, and training-data construction. An asset can look plausible in RGB while hiding disconnected geometry, problematic UV organization, or materials that fail to resolve correctly after import. We introduce Diagnostic Rendered Evidence (DRE): aligned beauty, wireframe, and UV-seam-plus-wireframe views scored by pass-specific frozen ViT classifiers and combined through late fusion. On 1,234 human labeled TexVerse assets, nested grouped evaluation achieves 86.36% balanced accuracy and 88.32% bad-asset recall. On a separate, balanced 764 external reference test, DRE achieved 88.9% balanced reference-label accuracy. Our method requires only an imported asset and renderer-derived evidence, without repository identity, creator, popularity, text, or source metadata at inference. Our results support cross-corpus portability and show that purpose built diagnostic views expose defects that just the beauty views can miss. Webpage: https://rbz-99.github.io/is-this-3d-asset-usable/

arXiv:2611.04101v1cs.AI

Structured Clinical Memory as Context for Probabilistic Decision Models: A Synthetic Evaluation with Jev

Longitudinal clinical context is often distributed across noisy, repetitive forms and narrative records. We study whether a compact structured memory can improve a probabilistic decision model's answers when the underlying patient vignette is already available. We created 30 fully synthetic cases spanning twelve clinical domains, paired each with synthetically generated structured memory, and evaluated 120 typed questions with Jev. The raw-vignette arm achieved 90.0% accuracy (95% normal-approximation confidence interval: 84.6%-95.4%); adding memory achieved 93.3% (88.9%-97.8%). Five questions changed from incorrect to correct and one changed in the opposite direction (exact McNemar $p=0.219$). Choice log loss decreased from 0.258 to 0.175 and binary Brier score from 0.054 to 0.022. Memory added 310 input tokens per call on average, while observed mean latency was unchanged at approximately 0.5 seconds. These results are directionally positive rather than clinically validating: the cases are synthetic, the labels are single-author judgments, and the memory is an oracle upper bound rather than extracted production output. The pilot supports evaluating structured clinical memory as a reusable state interface for decision models.

arXiv:2611.04100v1cs.AI

The Agent Board: Kanban for Machines; Durable Coordination for Agents That Outlive the Conversation

Harnesses for large language models have converged on a single-threaded shape: one agent, one history, one machine, one sitting. That shape fails work that spans many agents, outlives any one model call, or waits on systems that are not agents. We describe the Agent Board, the coordination substrate in the Corvic harness. The board is a durable, shared table of semi-semantic tasks: a few exact fields the system routes on, and a free-form body that only language models read. Agents post tasks, watch tasks, and block on them, ending their turn and holding no process until a matching post wakes them. A system agent, the Board Keeper, acts as the interrupt handler for the board: it carries events from non-agent parts of the harness onto the board, matches them to waiting tasks by reading their prose, and allocates new agents for tasks that name a kind of agent rather than an instance. Loadable coordination policy kernels give statuses and roles meaning for a given pattern of work, without the board enforcing them. The design mirrors the kanban board a human team puts on a wall. We argue it is what persistent agent products such as Meta Muse, OpenAI dots and SpaceXAI Grok Bot need once an agent must wait, hand off work, or work with agents from another vendor. We view the board as an operating-system scheduler for agents and state where the analogy breaks. We separate what the mechanism guarantees from what it leaves to convention, and report a trace from a live board and three small simulated pilots. In the pilots, the wait cost of the board was flat in wait length while naive polling grew about 165-fold, but a plain external watcher was cheaper than the board. The Board Keeper matched prose waiters with no false wakes but missed 3.6 percent of real wakes.

arXiv:2611.04099v1cs.CL

Evaluation Contracts and Evaluation Leverage: Sliced-Gap Readings in Structured Prediction Can Move Without Model Change

NLP evaluation reports routinely cite sliced performance differences, most prominently the intra-inter gap in document-level relation extraction, as capability evidence. We show that such gap readings are joint properties of models and evaluation contracts: scoring function, gold version, threshold, output format, and prompt language. With predictions frozen, a method ordering reverses under a gold-standard exchange (+0.0363, CI excluding zero; tie-level after), movement localizing where gold re-partitions: renames are gauge transformations (Delta=0; re-annotation +0.862). We formalize this as an evaluation leverage account centered on the rank-flip condition. Constructively, the audited evaluation reading (AER) is a contract-marginalized, tie-declared, seed-prescribed, gate-filtered procedure whose tie declarations cover every contract-induced flip on the declared WMT19 grids (14/14; held-out 73% vs. union-of-ties 89%), matching the best single metric contract at epsilon=0.10 equivalence. Evidence spans DocRE, HacRED, a UD probe, and released gold revisions; an alternative-reference-set replication extends it beyond our testbeds. Failed predictions retained as boundaries; a 5% violation gate flags unverifiable settings and bounds survivor-selection bias. We release EvalLever, a preregistered workflow with an amendment trail; on LLM-as-judge grids, the audited ensemble stabilizes every unstable pair (13-15 to zero; generic ensemble shrinkage: the preregistered correlation criterion failed, the companion stabilization criterion passed).

arXiv:2611.04098v1cs.LG

When Strong Reasoners Join the Pool: Small Union Premiums and Costly Runaway Generations in LLM Routing

Routing benchmarks often judge a router against a hindsight oracle that counts a query as solved whenever any model's recorded answer is correct, but under stochastic decoding that oracle mixes two different objects: the clairvoyant ceiling of committing to one model per query and the premium of calling several models and keeping any correct answer. A companion study measured this premium at 3.5 percentage points (pp) for 11 small open models on MATH-500; how large it is once current reasoning models enter the pool was unknown. We measure it with a fixed-protocol, pre-committed generation on a single NVIDIA B300: all 500 MATH-500 problems, 30 seeded samples per problem, Llama-3.3-70B-Instruct and NVIDIA's Nemotron Nano 9B v2 and Nemotron Super 49B v1.5 (a NAS derivative of the same 70B) under vendor sampling, a 65,536-token cap, a pinned Math-Verify scorer, and an append-only attempt ledger, for 45,000 scored responses. The two reasoning models both reach 96.1% accuracy against 76.2% for the 70B baseline. The empirical product-coupling premium is 0.215-0.588 pp across all two- and three-model pools; the empirical plug-in selection headroom over always calling Nano is 1.27 pp; and resampling Nano three times gives practically the same empirical any-correct rate as calling all three models once. Only 31 responses reached the cap, 25 of them with exactly periodic tails, yet the capped responses account for up to 60.2% of a model's output-length-squared proxy, and long generations made a fourth planned model, Llama-3.1-8B, too slow to finish. The premium is far below the companion's 3.5 pp, but models, pool size and protocol changed together; an oracle gap should therefore be reported with the pool, the protocol, the samples per cell and the decoding budget that produced it.

arXiv:2611.04097v1cs.LG

ScaleTraj: Scale-Decomposed Human Trajectory Generation

Individual-level trajectories are essential for applications that rely on user-level mobility behavior. Existing mobility-generation approaches face a fundamental tension between individual-level and population-level statistical fidelity: top-down models reproduce aggregate mobility statistics but struggle to guarantee realistic individual trajectories, while bottom-up generators synthesize individual trajectories yet often fail to preserve key population-level mobility regularities after aggregation. In particular, many bottom-up generated populations fail to reproduce the displacement distribution, whose heavy tail is well summarized by a truncated power law. We formalize this phenomenon as the \textbf{Population-Level Displacement Gap}. To investigate its prevalence, we conduct a systematic evaluation of representative model-driven and data-driven trajectory generators. Our results provide evidence that the Population-Level Displacement Gap is prevalent across diverse generation paradigms. To diagnose this gap, we introduce the \textbf{Scale Fidelity Profile (SFP)}, a scale-decomposed metric that quantifies discrepancies in trip allocations across spatial scales. SFP indicates that population-level displacement mismatches co-occur with systematic errors in how trips are allocated across spatial scales. Motivated by this observation, we develop \textbf{ScaleTraj}, an autoregressive Transformer that decomposes location prediction into a Scale Selection Head that decides the movement scale and a Scale-Aware Cell Selection Head that predicts the destination conditioned on the selected scale. Experiments on two datasets show that ScaleTraj reduces the Population-Level Displacement Gap versus model-driven and data-driven baselines, while preserving fidelity. Our code is available at https://github.com/SCU-TrustworthyAI/ScaleTraj.

arXiv:2611.04092v1cs.CV

CHROMA-MILL: Joint Low-Light Enhancement and White Balance under Colored Illumination

Capturing an image in a low-light environment produces a weak, noisy sensor signal. Low-light image enhancement (LLIE) aims to restore it. Auto white balance (AWB) is also required for every image, and the same noisy signal degrades its ability to infer the scene illuminant. LLIE is typically performed first, followed by AWB, but this strategy remains largely unexamined because existing low-light datasets span a narrow range of illuminants and lack illuminant ground truth for evaluating AWB. We address this with CHROMA-MILL, a RAW dataset of 50 scenes captured with a DSLR and smartphone under four calibrated illuminants at eleven intensity levels. Because light is additive, we blend these captures in the linear RAW domain to synthesize physically valid illuminant mixtures with exact chromaticity ground truth. Benchmarking on this data, we find joint correction is best, and introduce CTI-Retinexformer, which supervises an illumination map, a chromaticity prediction, and an illumination-invariant scene embedding, outperforming the best cascade by 0.7 dB PSNR$_L$ and 1.1 $\Delta E_{00}$.

arXiv:2611.04091v1cs.CL

Same Rule, Different Verdicts: Rubric Language and Clause Structure in LLM Judges for Korean Support Dialogues

Services that operate in Korean often write their evaluation rubrics in Korean and hand them to an LLM judge. We test whether a judge's verdict depends on the language of the rubric and on how a rubric item is phrased. We write 12 criteria for support dialogues in four clause structures (affirmative, negated, conditional, exception) such that two pairs of structures have the same meaning, and we give each item in Korean and in English. Six low-cost models from three providers, plus one reference model, judge 60 translated ESConv dialogues and 144 constructed boundary cases under every combination (51,744 requests across the seven models). The analysis plan was fixed and committed before data collection. Cross-provider agreement is lower with Korean rubrics than with English rubrics (Cohen's kappa 0.687 vs. 0.745; difference -0.058, 95% CI [-0.072,-0.045]). Disagreement concentrates in negated, conditional, and exception items (odds ratio 5.6 relative to affirmative items). A single model changes its verdict between two items with the same meaning in 11-15% of cases. Two hypotheses are not supported: models from the same provider do not agree more than models from different providers, and adding paired examples does not reduce the language gap. We measure consistency only; no human labels are used, and all materials were drafted by an LLM.

arXiv:2611.04090v1cs.CL

Swapping the Judge Changes the Ranking: A Preregistered Study of Cross-Provider Replication in LLM-Judged System Comparisons

Practitioners who compare systems with an LLM judge and have no human labels often check a conclusion by repeating it with a judge from another provider. We test how far that check can be trusted. In a preregistered study (with a disclosed post-registration reduction of the judge set), five judge models from three providers re-judged 4,000 Chatbot Arena battles (nine systems, 33 system pairs) that carry human votes, each in both response orders. Agreement with the human vote was low at the verdict level (median Cohen's kappa = 0.22), and system rankings did not survive a change of judge: Kendall's tau with the human ranking ranged from 0.17 to 0.89 (median 0.67), so the preregistered hypothesis that rankings remain stable was not supported. Conclusions replicated by two low-cost judges from different providers matched the human conclusion more often than single-judge conclusions (+0.10 precision), but the gain came from filtering out one poorly aligned judge, and replication involving that judge left a conclusion for only 35% of comparisons. Judges that agreed with humans reached a stable conclusion with no more verdicts than human votes required; the others failed to converge on 26-70% of system pairs at the sample sizes tested (up to 100 verdicts per pair). Disagreements with the human vote co-occurred across judges far more often than independence predicts, although a single human vote per battle cannot separate shared judge bias from voter noise.

arXiv:2611.04087v1cs.CV

ScienceVision: Can AI Observe and Reason Like a Scientist?

Scientific discovery builds on accurate observations interpreted through domain knowledge and reasoning. For scientific 3D data, this requires understanding structures and spatial relationships, from protein contacts to anatomical organization. Existing multimodal benchmarks largely evaluate text and 2D images, while scientific 3D evaluations vary in representations and protocols across disciplines, limiting cross-disciplinary comparison. To assess these foundational capabilities, we introduce ScienceVision, a cross-domain benchmark for scientific 3D understanding and reasoning through expert-level visual question answering (VQA). We construct 250 questions, with 50 each from Biology, Chemistry, Medical Imaging, Mechanical CAD, and Architecture, through a pipeline combining synthetic question generation with domain-expert review and annotation. Each task pairs a native 3D artifact with a scientific question, a specified response format, and a task-specific reference answer, requiring domain knowledge to be applied to geometric or topological evidence. We evaluate 12 models on the same questions in three settings: visual question answering with native 3D artifacts accessed through an agent harness's file and code-execution tools; the same workflow augmented with expert-inspired viewer inspection and measurement tools exposed through the Model Context Protocol (MCP); and visual question answering from prescribed rendered 2D images using an agent harness. All settings use the same task-specific deterministic final-answer scoring. The inspection environment supports view changes, target isolation, and slice inspection. In this main comparison, the best mean answer score is 0.53 out of 1. Mean scores in the native-artifact workflow without additional MCP tools exceed those in the rendered-image workflow for every model, by 17.2 percentage points on average.

arXiv:2611.04084v1cs.CV

From Visual Matching to Textual Conditioning: Cross-Domain Few-Shot Segmentation via Reference Inversion

Cross-Domain Few-Shot Segmentation (CD-FSS) aims to segment objects from unseen domains using only a few annotated references. Recent Segment Anything Model (SAM)-based methods have advanced CD-FSS, yet their reference-target visual matching is highly sensitive to target appearance variations. SAM3 introduces text-conditioned segmentation, enabling natural-language descriptions to guide segmentation across target images, but directly using the category name as the condition still yields unstable performance under severe domain shifts. We propose Reference Inversion for SAM3 (RI-SAM3) that leverages SAM3's large-scale vision-language pre-training to learn suitable textual tokens for each target category. RI-SAM3 comprises three core designs: (i) Bi-level Reference Inversion that learns complementary tokens before and after the frozen text encoder: the former remain anchored in SAM3's text space while the latter represent the category more flexibly; (ii) Layer-adaptive Textual Conditioning combines these tokens differently as object queries evolve across decoder layers; and (iii) Presence-aware Confidence Supervision that jointly supervises query confidence and target presence for reliable predictions. With SAM3 frozen, RI-SAM3 optimizes around 12K parameters only per category and it efficiently segments each target in a single forward pass with reusable textual conditioning without per-sample reference-target visual matching. Extensive experiments over multiple widely adopted benchmarks show that RI-SAM3 achieves superior performance consistently in CD-FSS tasks, together with strong robustness, efficiency, and generalization.

arXiv:2611.04079v1cs.AI

AlgoPilot: Cross-Paradigm Reasoning in Language Models via Strategy Selection and Guidance

Real-world problems span diverse domains, including program synthesis, symbolic reasoning, planning, and optimization, and often require fundamentally different solution paradigms. A central challenge for LLM-based reasoning is twofold: identifying an appropriate problem-solving approach and formulating it correctly. In this paper, we propose AlgoPilot, a cross-paradigm reasoning framework that enables adaptive selection of the solving approaches and structured problem formulation prior to execution. The AlgoPilot framework contains three components: 1) a SteerLM, trained via supervised fine-tuning (SFT) and reinforcement learning (RL), that adaptively selects the solving strategy (paradigm and modes) based on input problems, 2) the corresponding expert GuideLM, trained via SFT, that generates structured formulation guidance tailored to the selected strategy, and 3) an ExecutionLM that generates solutions conditioned on both the selected strategy and its formulation. Unlike prompting-based methods that rely on fixed reasoning templates, our approach learns to adaptively choose and structure problem-solving strategies based on the input. We build AlgoPilot on Qwen3-8B and evaluate across 5 problem categories and 13 benchmarks with 66 unique domains, comparing against both the latest similarly scaled small models and large LLMs. AlgoPilot significantly improves average accuracy for the same model from 38.4% to 72.9%, outperforming the best large LLM prompting baseline GLM-5 by 11.7%. Ablation studies confirm the effectiveness of each component. Notably, we show that the learned steering and formulation modules can transfer across execution models, suggesting their potential to improve other LLMs. Project page: https://sites.google.com/view/algopilot.

arXiv:2611.04073v1cs.LG

Computer-Use Agents Gain Little Benefit from a World Model When Conditioned on Tutorials

Recent work has improved computer-use agents (CUAs) by combining retrieved task-relevant tutorials and look-ahead planning using world models with MLLM judges. However, the contribution of each component -- tutorials, the judge, and the world model -- to overall improvement has not been evaluated in isolation. To fill this gap, we first construct OSWorld-Tutorial and ScienceBoard-Tutorial: two datasets for CUAs with curated multimodal tutorials, covering everyday and scientific domains. We then propose RAG-PolicyJudge (RAG-PJ), a policy-judge framework for CUAs conditioned on task-specific tutorials: the policy proposes candidate actions, and the judge selects among them. To isolate the effect of world models given tutorials, we compare RAG-PJ with and without common world models: using either zero-shot text-based descriptions of next states or an oracle that returns ground-truth visual observations. We find that RAG-PJ gains little benefit from these rollouts when tutorials are present, even with perfect visual futures. This suggests prior world-model gains in tutorial-conditioned methods come mainly from tutorial understanding and the judge's ability to select the proposed actions, rather than from predicted rollouts. Error analysis reveals that the bottleneck is not rollout fidelity, but the judge's ability to reason over future states and recognize task completion. This highlights the need to improve not just next-state prediction in world models, but also reasoning over their predicted rollouts. Datasets and code are available at https://github.com/realjoenguyen/RAGPJ-official.