Search arXiv⌕ Search

arXiv subjects

Nanyun Peng

Publications and source records attributed to Nanyun Peng.

At least 19 recordsLinked to original sources
arXiv2026

From Solvers to Research: Large Language Model-Driven Formal Mathematics at the Research Frontier

Recent developments in AI for Mathematics (AI4Math), especially Large Language Model (LLM)-driven theorem provers, has achieved remarkable success in formal proof generation for well-defined mathematical problems through Interactive Theorem Proving (ITP) languages. However, current systems remain fundamentally limited in tackling frontier research mathematics, such as discovering new theorems or resolving open conjectures, which are often open-ended, under-specified, and involve multiple layers of abstraction. We argue that the next leap in AI4Math systems requires a decisive shift from predefined problem-solvers to research agents that can address frontier mathematical challenges with rigorous formal mathematical reasoning. In this position paper, we provide a systematic review of the field, covering datasets, auto-formalization, and proof synthesis. More importantly, we identify core limitations of existing systems in serving as mathematical research agents, examining issues across datasets, relational structure, mathematical exploration, tool ecosystem, and human-AI collaboration, outlining a strategic road-map for the future of AI4Math.

cs.CL↗
arXiv2026

Diffusion Language Models Change Speed Across Tasks and Languages

Diffusion language models can write several tokens in one forward pass, and they decide during generation how many passes to spend. Two outputs of the same token length can therefore vary a lot in cost. To understand the underlying factors of this cost variation, we analyze two diffusion models across 10 tasks. We find that the number of forward passes per token depends heavily on the task. Text copied or lightly edited from the input is cheap, while open-ended writing is costly. In DiffusionGemma, generating a token for story writing requires more than three times as many forward passes as for grammar correction. Two output properties explain much of this gap: how much of it is new relative to the input, and how rare its words are within the task. The same holds within a single response when thinking is involved. The final answer is cheap when it repeats the model's reasoning trace and costly when it adds something new. In multilingual generation, we find that highly tokenized languages also require more passes per token, adding to the cost they already pay with more tokens. Finally, cost tracks correctness: on the same math problem, wrong answers take more passes per token than correct ones, and weighting candidate answers by decoding speed beats majority voting by up to 4.2 points at no extra cost. Taken together, our results show that diffusion's computational advantage depends on how much of the output is already determined, while its decoding effort offers a practical signal for making inference more efficient and reliable.

cs.CL↗
arXiv2026

KEEP-Bench: Evaluating Evidence Preservation in Context Compression

Context compression improves long-context efficiency but may introduce faithfulness issues through an information bottleneck. Existing evaluations rely on QA metrics, but a reader can still answer correctly from an unfaithful compression, so these metrics underestimate such loss. To bridge this gap, we introduce KEEP-Bench, a diagnostic benchmark that directly assesses compression faithfulness with 4,102 cases across three challenging settings that require keeping evidence for plausible reference resolutions, explicit qualifiers, and minority perspectives. Experiments show that even with tailored prompting, the strongest models remain unreliable: despite context windows exceeding 100K tokens and inputs averaging only 3.1K tokens, the best open-weight model (31B) preserves all required evidence in only 66.8% of cases, and the best proprietary model in only 61.6%. Performance also varies sharply across evidence structures, so conclusions from limited structures may be biased. We further introduce a metric, Evidence Preservation Return (EPR), to relate evidence recall to inference FLOPs, and find that it peaks at moderate scales (4B-8B), where recall remains low. Additional output budget mainly benefits already strong models and widens capability gaps, yet the stronger preservation of these models comes at a steep computational cost. These findings highlight the importance for future work to investigate compressors with strong information preservation under a small computation budget, rather than relying on a large output budget or model scaling.

cs.CL↗
arXiv2026

OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory

Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled study on 22 benchmarks shows that simply increasing the inference budget of the base model yields only modest gains. Such tasks demand grounded perception, external knowledge, and numerical measurements obtained through heterogeneous tools, whose outputs must be made interpretable to the model and retained across reasoning steps. We introduce OpenVLHarness, a multimodal reasoning harness organized around a capability layer and persistent multimodal memory. The capability layer combines task-level tool interfaces with evidence rendering, providing consistent access to heterogeneous backends and converting their outputs into model-readable observations. Persistent multimodal memory retains the corresponding images, structured data, and web references as addressable artifacts throughout each query. A coding interface executes generated programs over stored data, enabling computations that combine results from multiple tools. Together, these mechanisms support an iterative process of acquiring, inspecting, and composing multimodal evidence. Evaluated on 22 benchmarks with eight open and proprietary backbones, OpenVLHarness improves average scores by 9.7-12.6 points over matched no-tool baselines, including +11.2 with Qwen3-VL-8B, +9.7 with Kimi K3, and +9.9 with GPT-6 Astra. With GPT-6 Luna at medium reasoning effort, OpenVLHarness outperforms the base model, Codex, and Visual Sketchpad at maximum effort by 7.7-8.5 points while incurring at most 37% of their model-token cost.

cs.CL↗
arXiv2026

Nexus : An Agentic Framework for Context-Aware Time Series Forecasting

Language models have emerged as highly capable reasoners, yet their performance in time series forecasting remains uneven across domains and levels of contextual grounding. This is primarily due to their suboptimal time series reasoning and numerical modeling capabilities. To tackle this, we introduce Nexus, a multimodal forecasting framework that bridges qualitative event reasoning with structural numerical modeling through principled statistical synthesis. Nexus follows three structured stages: (1) Temporal profiling and anchor generation, which distills the multimodal history into a timestamped causal-temporal profile and generates a forecast anchor (2) Dual contextual outlook generation, which produces decoupled macro and micro level forecast trajectories, and (3) Bayesian synthesis, which reconciles the forecasting anchor with both contextual forecast trajectories via closed-form Gaussian updating weighted by empirical error covariances. Evaluated across three diverse forecasting benchmarks constituting eight datasets spanning rich event contexts, abrupt shocks, and purely numerical series, Nexus consistently outperforms leading time-series foundation models and language model baselines, while providing interpretable causal reasoning.

cs.AI↗
arXiv2026

STRIDE: Evaluating Spatiotemporal Reasoning in Driving Edge Cases

In autonomous driving, perceiving surrounding objects is not enough: failing to understand how they move and interact over time can lead to dangerous decisions. Evaluating vision-language models (VLMs) therefore requires testing whether they connect past spatial observations to future consequences. Yet existing benchmarks often assess edge cases through basic perception questions or test temporal understanding in simple scenarios, leaving complex spatiotemporal reasoning underexplored. We introduce STRIDE (SpatioTemporal Reasoning In Driving Edge Cases), an object-centric benchmark for long-tail, safety-critical driving scenarios. STRIDE covers six task families: spatial perception, spatial understanding, temporal memory, temporal extrapolation, trajectory prediction, and scene-context awareness. It pairs multiple-choice questions probing individual spatiotemporal judgments with related free-form visual question answering (VQA) requiring detailed scene explanations and justifications of driving decisions. We prioritize scenes with unusual object behaviors and complex traffic interactions, generate reference answers from structured dataset annotations through automated procedures, and use limited language-model assistance and human verification to ensure annotation quality. Across 12 proprietary, open-weight, and driving-specialized models, even the best performer, GPT-6-Astra, achieves only 43.4% average multiple-choice accuracy, versus an approximately 20% random-choice baseline. Further analysis reveals that strong performance in describing driving scenes does not necessarily translate to accurate safety-critical judgments, and that models struggle more with reconstructing object motion than recalling past positions. STRIDE thus exposes spatiotemporal driving limitations that aggregate VQA scores can conceal.

cs.CV↗
arXiv2026

Lean Graph Search for Formal Theorem Proving

Retrieving relevant lemmas is crucial for formal theorem proving, enabling provers to build machine-checkable proofs from established mathematical results. Recent retrieval systems match library declarations to a target theorem or its intermediate proof steps, but make limited use of multi-hop dependencies among declarations. These dependencies can reveal supporting lemmas that are not necessarily lexically similar to the target theorem but recur in proofs of related problems. To this end, we propose LeanGraphSearch, a graph-based retrieval method for Lean 4 that constructs a graph from Mathlib's forward dependencies and traverses it to uncover supporting lemmas through multi-hop dependency chains. Specifically, LeanGraphSearch links each declaration to the premises it depends on, uses a graph-based ranking algorithm to propagate relevance from semantically matched declarations to their dependencies, and reranks the expanded candidates against the original query. Across two mathematical retrieval benchmarks, LeanGraphSearch improves recall by up to 12% over the strongest baseline, LeanSearch v2, on MathlibQR and increases coverage of complete annotated proof routes on MathlibMPR, covering up to 3.7% of extra groups. More importantly, we evaluate whether these retrieval gains translate into improved end-to-end theorem proving by integrating LeanGraphSearch into an iterative proving loop guided by Lean feedback. Across three theorem proving benchmarks, LeanGraphSearch improves pass rate at {8, 16, 32} reflection turns by as much as 40% compared to vanilla prover and at most 10% compared to LeanSearch v2, showing higher accuracy-cost efficiency. These findings highlight the value of dependency graph search for theorem proving and motivate further research on structure-aware search for proof discovery.

cs.LG↗
arXiv2026

Geometric Calibration of Supervised Fine-Tuning

Recent studies investigate the weight geometry through its spectral structure, which characterizes how the weights transform different input directions. Within this view, supervised fine-tuning (SFT) is shown to induce dense updates that perturb the spectral geometry of pre-SFT weights along their principal directions, potentially disrupting the model's prior knowledge. These observations motivate using the geometry of pre-SFT weights as a reference for refining learned SFT updates. We study this perspective as Geometric Calibration and develop SPACE Subspace Projection for Adapted Checkpoint Enhancement), a lightweight post-hoc method that selectively reverts SFT update components within the principal subspace of the pre-SFT weights while preserving those in its orthogonal complement. SPACE is a data-free, training-free method that requires only the pre- and post-SFT checkpoints. We evaluate SPACE on Qwen2.5, Qwen3, Qwen3.5, Llama-3.1, and Olmo-3 models ranging from 2B to 14B, across mathematical, coding, and medical SFT, as well as several SFT variants. Overall, SPACE mitigates out-of-domain (OOD) capability regression while further improving in-domain (ID) performance over the uncalibrated SFT checkpoints. For math SFT, SPACE yields average ID/OOD gains of 2.85/2.26 points on Qwen2.5-7B, and 2.41/1.52 points on Qwen3-14B-Base, respectively. These results highlight how the geometric relationship between SFT updates and pre-SFT weights shapes the adapted model's capabilities and, more broadly, motivate geometry-aware parameter updates for continual model improvement.

cs.LG↗
arXiv2026

Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards

Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely for lack of both training data and a working recipe, and scaling both is hard: long, complex tasks are difficult to curate and verify, and their rewards are sparse. We address both. First, we introduce Tev-Gym, the first long-horizon agent training dataset with verifiable, dense, progressive rewards -- 1,170 verified task environments across 7 domains. It is built on a latent-world-first approach that first constructs a latent world: a complete and correct hidden specification of the task domain, workflow, and expected state after each step. Each task is then progressively rendered from this specification into an agent-facing environment, enabling controllable complexity and a natural progress metric. Second, we turn this metric into a training recipe: via progress-potential reward shaping (PPRS), we convert progress into a dense per-step reward that improves training efficiency and makes otherwise-intractable tasks trainable. Training with this recipe yields Tev-27B, which achieves substantial gains on the Agents' Last Exam computing and math sciences domains (ALE-CM), matching GPT-6 Astra and Claude Opus 5 at Pass@5. Despite training only on ALE-related data, it generalizes to the Terminal-Bench (TB) 3.0 and 4.0 subsets, surpassing GLM 5.2 and Gemini 3.7 Flash, and reaches Claude Opus 4.8-level performance on middle-horizon tasks such as TB 2.1.

cs.CL↗
arXiv2026

DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck

Evaluation has been central to advances in AI, and it is equally critical to autonomous discovery. Yet scaling autonomous discovery to real-world engineering challenges faces a fundamental bottleneck: the cost of evaluating progress. In real settings, a single true evaluation, such as a full benchmark or serving trace, can take hours, so a search can afford only a handful of true evaluations. We present DuoDiscover, a search system that jointly discovers what to optimize and how to evaluate progress efficiently, through three mechanisms. First, each worker in a fleet of coding agents co-discovers an optimization target and a fast, tailored proxy evaluator, which we call a yardstick. The agent then hill-climbs the yardstick rather than the true evaluation. Second, since search can exploit gaps between a yardstick and the true objective, fidelity tests continually recheck each yardstick's ranking and invalidate any that drift. Third, the fleet is bottom-up, with no central decomposition of the problem. A lightweight orchestrator spawns and terminates agents, briefs running ones, and seeds new agents with others' confirmed work. DuoDiscover enables autonomous optimization from software to chip architecture. We demonstrate DuoDiscover on two repository-level benchmarks and four GPU and TPU kernel suites covering hundreds of kernels. On SWE-fficiency, it doubles the average speedup of human experts. On SOL-Exec, it ranks first on 121 of 235 kernels and in the top three on 212. On the heavily optimized vLLM and SQLite engines, DuoDiscover achieves 1.92$\times$ and 1.38$\times$ gains over the base code. On a CPU microarchitecture optimization task with evaluations of up to 28 hours, DuoDiscover surpasses both AlphaEvolve-based ArchAgent v2 and the best human design, with 1.65$\times$ ArchAgent v2's gain in 16.7 days versus more than 40 days, advancing the frontier of autonomous chip design.

cs.CL↗
arXiv2026

VIS-Ground: Video Interactive Storytelling with Contextual Grounding

Video interactive storytelling enables viewers to actively steer how a video unfolds. However, once we allow viewers to intervene during generation, a new challenge arises: The viewer's request can have latent dependencies on both the grounding source and the current rendered video state. These dependencies may not be explicitly stated in any individual input, but emerge only when the source, rendered history, and new viewer intent are considered jointly. Existing interactive video generation systems primarily emphasize following viewer instructions, while source-grounded video generation methods focus on aligning generated content with an external narrative or knowledge source. This leaves a fundamental question underexplored: What context should a generation model ground on during interactive continuation, and how can heterogeneous, unstructured inputs be transformed into such grounding context? In this work, we formulate contextual grounding as the process of transforming heterogeneous input context into an executable constraint model for video generation. To address this challenge, we introduce VIS-Ground, which performs Structured Context Abstraction to recover grounded states and cross-context dependencies, Generation Constraints Induction to project relevant dependencies into candidate-specific constraints, and Constrained Video Generation to enforce these constraints through planning, verification, revision, and rendering. Across three video generation backbones, VIS-Ground consistently achieves the highest overall composite score, reaching an average absolute improvement of 10.3 points over the strongest per-backbone baselines. Detailed analysis further shows gains across both narrative and knowledge grounding, and reveals remaining challenges in dependency extraction, and faithful realization during video rendering.

cs.CV↗
arXiv2026

Cross-Lingual Alignment for Decoder-Only Models using MoE Routers

Cross-lingual contrastive learning has been a core component of multilingual encoder training, but the ability to explicitly align representations is not possible in decoder-only LLMs because of varying multilingual tokenization. However, a growing amount of research suggests that even in LLMs, higher cross-lingual representational alignment leads to improved cross-lingual transfer. In this paper, we propose a novel approach to reimagine cross-lingual contrastive learning given the architectural constraints of modern LLMs. Rather than applying an auxiliary alignment loss on hidden states, we propose using the outputs of the mixture-of-experts (MoE) routers as the target for alignment. Router outputs lend themselves better to pooling over many tokens, enabling more reliable cross-lingual comparisons at the sequence-level. Controlled continual pre-training experiments on four open-source MoEs show that incorporating this routing loss also aligns the underlying hidden representations across languages. Most importantly, this loss improves multilingual performance on our diverse evaluation suite, demonstrating the potential of cross-lingual MoE router alignment.

cs.CL↗
arXiv2026

Continual Graph Memory for Mathematical Research Agents

Using frontier agent harnesses to tackle mathematical research problems has emerged as an effective means of advancing mathematics. However, solving frontier problems in mathematics may require a massive number of agents working in parallel for extended periods to construct proofs, thereby generating an enormous volume of intermediate proof results. Organizing these intermediate results throughout a long-horizon proof-search process and reusing knowledge gained from prior explorations remain major challenges. We present Ansatz, a mathematical research agent built around Continual Graph Memory, a graph-based, evolvable, cross-problem mathematical research memory system that explicitly organizes the entire proof search process and reuses information from exploration trajectories of previous problems. Specifically, we develop a unified graph memory that represents all intermediate exploration results, including facts, plans, and counterexamples, together with edges that explicitly represent the relationships among them; dependency-aware retrieval supplies precisely targeted local context; an evidence-sensitive curator updates the research frontier and distills lessons from prior attempts; and scoped recall surfaces earlier statements and negative findings for local re-proving rather than uncritical reuse. Experiments cover runs across all ten First Proof Second Batch problems, together with four component studies. Ansatz reports closure on all ten research tasks, demonstrating its ability to sustain and resume long-horizon mathematical search. Beyond these problems, Ansatz also produces solutions to the Jamison caterpillar conjecture and Erdős Problems 289, 348, and 488 without human intervention, and makes partial progress on several open problems, illustrating its strong ability to solve open mathematical research problems.

cs.AI↗
arXiv2026

SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting

Real-world time series are frequently driven by exogenous events and structural shifts, rendering conventional forecasting based solely on historical numerical observations insufficient. While language models can retrieve external news, standard retrieval-augmented approaches struggle with high noise, missing signals, and an inability to reason causally about event impacts. We propose SEER (Self-Evolving Event Reasoning and Retrieval), a closed-loop framework that dynamically optimizes event conditioning for time series forecasting. SEER translates prediction errors into two decoupled feedback mechanisms: (i) a reflective retrieval memory that refines subsequent search queries and filters spurious noise, and (ii) a persistent causal knowledge base that distills transferable domain dynamics. SEER enforces strict chronological boundaries across both event retrieval and reflection, preventing look-ahead bias and data leakage. Across six volatile time-series benchmarks, SEER consistently outperforms state-of-the-art time series foundation models and language model baselines.

cs.LG↗
arXiv2026

LEAF: A Living Benchmark for Event-Augmented Forecasting

Large Language Models (LLMs) are increasingly applied to real-world forecasting tasks, yet evaluating their true predictive capability remains compromised by pre-training data contamination and look-ahead leakage in automated search. Existing benchmarks either rely on static contexts, restrict evaluations to narrow environments, or fail to audit auxiliary textual events for future information leakage. To establish a rigorous evaluation paradigm, we propose LEAF, the first living benchmark for event-augmented forecasting tasks, including trend, event, and time series forecasting. LEAF couples a recursive retrieval agent system with dual-agent cross-validation to gather comprehensive, relevant, and temporally aligned auxiliary context. A comprehensive audit across 500 tasks by 47 domain specialists demonstrates that our pipeline suppresses future information leakage from 8.6% to 1.6%. Across extensive evaluations of 16 frontier proprietary and open-weight models on our benchmark, we show that LLMs effectively extract signals from verified events to boost trend and event forecasting.

cs.LG↗
arXiv2026

From Routing Signals to Selective Review: Visual regrounding in MoE VLMs

Vision-language models (VLMs) may accept false visual premises, answering questions about a target object's color, count, location, or state even when it is absent. We call this reliability-critical behavior a target-absence grounding failure. Existing visual-grounding detectors primarily rely on generated responses, hidden states, or uncertainty measures. We present the first framework to leverage internal routing decisions in Mixture-of-Experts (MoE) VLMs to detect target absence before generation and guide selective correction. We extract target-token routing probabilities from Qwen3-VL-30B-A3B-Instruct and Gemma-4-26B-A4B-it, train a separate L2-regularized linear detector for each model, and use its predictions to selectively invoke a target-aware review prompt. Using routing alone, the Qwen and Gemma detectors achieve ROC-AUCs of 0.9988 and 0.9956 on GQA-Inpaint and retain 0.8095 and 0.7781 on the external OBER dataset, respectively. The resulting routing-gated policy improves end-to-end accuracy on GQA-Inpaint and OBER by +22.25% and +12.17% for Qwen, and by +13.42% and +1.39% for Gemma, without modifying model weights. Further analysis shows that the signal is localized to the target-object token, emerges in early MoE layers, and is distributed across partially substitutable experts. Although cross-dataset threshold shifts require recalibration, false-positive review causes limited harm overall, suggesting that intervention risk can be controlled through joint selection of the detector threshold and review prompt. Overall, we show that routing probabilities alone preserve actionable information about visual perception, allowing computation already produced by an MoE VLM to support low-cost detection and selective visual regrounding.

cs.CL↗
arXiv2026

Multilinguality in Hybrid Attention LLMs

In response to the growing demand for long sequences in agentic and reasoning use cases, many state-of-the-art LLMs combine multiple variants of attention to mitigate the quadratic complexity of traditional softmax attention. These hybrid attention LLMs aim to balance the strengths and limitations of full attention and alternatives based on recurrence. This work presents a first study of how hybrid attention impacts the multilinguality of LLMs. Beyond the impact on long sequences in poorly tokenized languages, our study is motivated by the possibility that the inductive biases of the recurrent state alter linguistic processing. Our interpretability analysis confirms this, showing that cross-lingual representations in hybrid models develop in patterns tied to the ordering of recurrent and full-attention layers. Across diverse models, we notably observe a pronounced spike in cross-lingual alignment around the first full-attention layer. These findings lead us to question the conventional ordering of attention layers. In distillation experiments on multilingual data, all alternative layer orderings outperform the standard throughout training, learning up to 2.5X faster. These stark, replicable results prompt our theory that multilingual models would benefit from starting with a full-attention layer rather than recurrent layers.

cs.CL↗
arXiv2026

Overwhelmed by Choice: Studying LLM Decision Making at Scale

Multiple-choice and candidate-selection evaluations are widely used to assess LLM reasoning and decision-making, yet most benchmarks contain relatively small candidate sets. It remains unclear whether conclusions drawn from these settings remain valid as the candidate space scales. We systematically evaluate LLMs as the number of competing candidates increases and find substantial accuracy degradation across tasks, prompting strategies, and model scales. Controlled analyses show that standard long-context retrieval explanations cannot fully account for this degradation. Instead, we identify two systematic failure patterns. First, gold-margin collapse: the score gap between the correct answer and the strongest distractor progressively shrinks, driven primarily by weakening confidence in the correct answer. Second, earlier candidate preferences become increasingly difficult to overturn, with later candidates exerting progressively weaker influence on the final prediction. Motivated by these findings, we evaluate hierarchical partitioning and permutation-based inference, which improve accuracy by roughly 20 percentage points at $N=160$ on both HotpotQA and MIMIC. Overall, our results identify candidate-set scale as an important evaluation-protocol variable and show that strong small-option performance does not necessarily imply robust large-scale candidate comparison.

cs.AI↗