A study evaluated several AI systems on ten research-level mathematics problems created by prominent mathematicians. The results include AI-generated solutions, human solutions, and referee reports, offering a detailed assessment of AI performance in solving advanced mathematical problems.
Language models of GPT-2 size cannot independently discover zero during testing, regardless of pretraining. However, performance improves significantly with training on tens to hundreds of zero examples, and language pretraining reduces required examples by about 50%.
As enterprise agent tool catalogs scale from 10 to 110 agents, routing accuracy drops 16--23 percentage points on under-specified requests. An oracle analysis identifies retrieval and confusion gaps, with embedding-based shortlisting recovering +10--11pp F1. A human-annotated study of 1,435 utterances confirms real-world recovery of +10--17pp despite lower absolute performance.
A new framework uses prompt perturbation to identify and filter structurally inconsistent pairwise comparisons in large language model evaluations. By incorporating graph-level consistency checks before ranking aggregation, the method reduces cyclic preferences and improves the reliability of LLM rankings.
We present a framework for evaluating agentic skills by constructing realistic tasks and assessing skill utility through task execution. Applied to 500 real-world skills, it generates 1,000 tasks and scoring rubrics, evaluating 19 agent-model configurations across proprietary and open-source models. Results show significant variation in instruction adherence and performance gains, with skills substantially altering model behavior compared to no-skill setups.
A study finds bilingual fine-tuning enhances automatic speech recognition in low-resource languages when language identification is accurate. Including a language identification token at inference improves ASR performance when identification accuracy is low, especially in diverse language pairs across different families and writing systems.
NNN decoding selects documents as a joint set that jointly reconstructs the query embedding via a sparse non-negative linear combination. It strictly extends dense retrieval by handling queries that dense retrieval fails on, especially in corpora with correlated documents, and achieves superior performance through end-to-end training of embeddings.
Glimmer 1 is a 10,000-parameter language model trained on 500K tokens from FineWeb-Edu. It features a 512-token context window, a standard Llama architecture with 16 hidden dimensions, 2 layers, 4 attention heads, and 1 KV head using GQA, and is available on Hugging Face.
A measurement study finds that 26 semantic post-hoc operators do not improve held-out accuracy over Best-of-N in frozen small code models. While two operators—expression-layer recovery and adaptive consensus early-stop—offer benefits in compute efficiency or program recovery, none outperform BoN in accuracy. The results highlight systemic limitations in error detection and coverage, suggesting that model harnesses and error coverage must be improved before post-hoc reasoning is considered.
IMPACTeen is a dataset of 1,021 texts annotated from five perspectives—teenagers, parents, psychologists, communication experts, and teachers. It includes 5,100 annotation records covering social influence techniques, intentions, consequences, and resistance, with annotations validated through human editing. The dataset, created using LLM generation and human validation, is available in both Polish and English and supports research on social influence and language model training.
MA-SBI introduces a calibration-free simulation-based inference framework that uses side-channel text, like regime labels or instructions, to correct for simulator misspecification. It employs a learned corrector to apply observation-space shifts before posterior inference, without needing ground-truth parameter pairs or retraining. On hide-the-calibration benchmarks, MA-SBI matches the oracle posterior with text alone, outperforming RoPE under limited data, and shows robustness on real-world epidemiological and cognitive-science datasets.
Analysis of 56,800 AI conference papers shows documentation practices improved from 2014 to 2024. Papers sharing both code and data increased from 11% to 64%, and estimated reproducibility rose from 28% to 64%. These improvements predate formal reproducibility checklists, indicating a broader shift toward open science.
Experiments on EC 2025's 'Stable Menus of Public Goods' show that human-intuition prompts improve LLM performance and multi-turn interactions enhance ambitious steps. However, when compared to a first-year PhD student using an unpublished manuscript, the LLM is found to be slightly less effective.
Public AI evaluation archives show that a single terminal result can arise from two distinct pre-terminal histories, with estimated times to reach 95% of performance ceilings at 23.03 or 75.13. A candidate selection-aware frontier model fails synthetic recovery and uncertainty calibration, and is rejected by fixed audit gates. An archive-and-adjudication protocol verifies timing boundaries and falsifies unsupported frontier claims.
HAMON uses passive optical components to perform long-horizon time-series forecasting, outperforming top digital models on ETTm2 across all horizons and on ETTh2 at all but the longest horizon. It achieves up to 14% lower MSE and relies on physical optical propagation without trainable digital layers, demonstrating that passive optical mixing can produce competitive forecasts.
The study introduces PReLU-IPM, a new integral probability metric based on a neural network discriminator with a single node. The resulting PReLU-TST test is nonparametric, consistent, and asymptotically equivalent to standard IPM-based tests, showing higher power or competitive performance on simulated and real datasets.
A new activation function enables fixed-size neural networks to approximate any function in Sobolev spaces $W^{s,\infty}((a,b)^d)$ with arbitrary accuracy in the $W^{s-1,\infty}$-norm. The results use elementary activations like EUAF and DUAF$_\infty$, with explicit width and depth bounds, and extend to sigmoidal variants $\widetilde{\mathrm{DUAF}}_n$ preserving accuracy for all $1\leq s\leq n$.
A compact spectral representation using Betti numbers, spectral gap, and analytic torsion distills persistent Laplacians into three mathematically grounded invariants. This approach captures essential predictive signals from the full spectrum, outperforms it in some cases, and reduces computational overhead on datasets like MNIST, QM-3D, and SKEMPI WT.
A new multi-center benchmark enables abdominal disease diagnosis and report generation from non-contrast CT by synthesizing contrast-enhanced findings. The dataset includes paired NCCT-CECT studies and reports from two centers, showing NCCT achieves average multi-organ AUCs of 69.1% internally and 63.1% externally. The benchmark and code are publicly released to support research into safer, contrast-free abdominal imaging workflows.
HAMON uses passive optical diffraction to generate forecasts, outperforming digital baselines on ETTm2 at all horizons and ETTh2 at all but the longest horizon. It achieves up to 14% lower MSE and operates without trainable digital mixing, relying instead on physical optical propagation.