arXiv subjects

Tianfu Fu

Publications and source records attributed to Tianfu Fu.

7 recordsLinked to original sources

arXiv2026

DeployBench: Benchmarking LLM Agents for Research Artifact Deployment

LLM agents have made rapid progress on software engineering and ML research tasks, but these advances often assume access to a working runnable environment. For research artifacts released alongside published papers, setting up such an environment from a fresh machine remains a major bottleneck. Existing environment setup benchmarks do not cover the full scope of research-artifact deployment, which involves multi-language toolchains, system-level dependencies beyond containers (e.g., GPU/CUDA and kernel configurations), and legacy artifact compatibility. We introduce DeployBench, a multi-domain benchmark of 51 research-artifact deployment tasks spanning AI/ML, computer systems, and scientific computing, covering all these dimensions. Each task is verified by a hidden pipeline that executes the paper's designated experiment and checks its outputs. Evaluating five state-of-the-art LLMs with OpenHands yields pass-rates from 7.8%-51.0%. Failures are dominated by a completion-judgment problem: 102 of 181 are agent-terminated self-stops, where the agent's pre-finish checks validate a different or weaker target than the paper-specific task requires. DeployBench highlights the gap between current agents and autonomous deployment, and offers a realistic testbed for scientific research agents.

cs.SE↗

arXiv2026

FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale

Many real-world coding challenges are open-ended and admit no known optimal solution. Yet, recent progress in LLM coding has focused on well-defined tasks such as feature implementation, bug fixing, and competitive programming. Open-ended coding remains a weak spot for LLMs, largely because open-ended training problems are scarce and expensive to construct. Our goal is to synthesize open-ended coding problems at scale to train stronger LLM coders. We introduce FrontierSmith, an automated system for iteratively evolving open-ended problems from existing closed-ended coding tasks. Starting from competitive programming problems, FrontierSmith generates candidate open-ended variants by changing the problems'goals, restricting outputs, and generalizing inputs. It then uses a quantitative idea divergence metric to select problems that elicit genuinely diverse approaches from different solvers. Agents then generate test cases and verifiers for the surviving candidates. On two open-ended coding benchmarks, training on our synthesized data yields substantial gains over the base models: Qwen3.5-9B improves by +8.82 score on FrontierCS and +306.36 (Elo-rating-based performance) on ALE-bench; Qwen3.5-27B improves by +12.12 and +309.12, respectively. The synthesized problems also make agents take more turns and use more tokens, similar to human-curated ones, suggesting that closed-ended seeds can be a practical starting point for long-horizon coding data.

cs.LG↗

arXiv2026

TriAttention: Efficient Long Reasoning with Trigonometric KV Compression

Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks. Leading KV cache compression methods estimate KV importance using attention scores from recent post-RoPE queries. However, queries rotate with position during RoPE, making representative queries very few, leading to poor top-key selection and unstable reasoning. To avoid this issue, we turn to the pre-RoPE space, where we observe that Q and K vectors are highly concentrated around fixed non-zero centers and remain stable across positions -- Q/K concentration. We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are preferred via a trigonometric series. Based on this, we propose TriAttention to estimate key importance by leveraging these centers. Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional signal for importance estimation. On AIME25 with 32K-token generation, TriAttention matches Full Attention reasoning accuracy while achieving 2.5x higher throughput or 10.7x KV memory reduction, whereas leading baselines achieve only about half the accuracy at the same efficiency. TriAttention enables OpenClaw deployment on a single consumer GPU, where long context would otherwise cause out-of-memory with Full Attention.

cs.CL↗

arXiv2026

Agent Banana: High-Fidelity Image Editing with Agentic Thinking and Tooling

We study instruction-based image editing under professional workflows and identify three persistent challenges: (i) editors often over-edit, modifying content beyond the user's intent; (ii) existing models are largely single-turn, while multi-turn edits can alter object faithfulness; and (iii) evaluation at around 1K resolution is misaligned with real workflows that often operate on ultra high-definition images (e.g., 4K). We propose Agent Banana, a hierarchical agentic planner-executor framework for high-fidelity, object-aware, deliberative editing. Agent Banana introduces two key mechanisms: (1) Context Folding, which compresses long interaction histories into structured memory for stable long-horizon control; and (2) Image Layer Decomposition, which performs localized layer-based edits to preserve non-target regions while enabling native-resolution outputs. To support rigorous evaluation, we build HDD-Bench, a high-definition, dialogue-based benchmark featuring verifiable stepwise targets and native 4K images (11.8M pixels) for diagnosing long-horizon failures. On HDD-Bench, Agent Banana achieves the best multi-turn consistency and background fidelity (e.g., IC 0.871, SSIM-OM 0.84, LPIPS-OM 0.12) while remaining competitive on instruction following, and also attains strong performance on standard single-turn editing benchmarks. We hope this work advances reliable, professional-grade agentic image editing and its integration into real workflows.

cs.CV↗

arXiv2026

Traversal-as-Policy: Log-Distilled Gated Behavior Trees as Externalized, Verifiable Policies for Safe, Robust, and Efficient Agents

Autonomous LLM agents fail because long-horizon policy remains implicit in model weights and transcripts, while safety is retrofitted post hoc. We propose Traversal-as-Policy: distill sandboxed OpenHands execution logs into a single executable Gated Behavior Tree (GBT) and treat tree traversal -- rather than unconstrained generation -- as the control policy whenever a task is in coverage. Each node encodes a state-conditioned action macro mined and merge-checked from successful trajectories; macros implicated by unsafe traces attach deterministic pre-execution gates over structured tool context and bounded history, updated under experience-grounded monotonicity so previously rejected unsafe contexts cannot be re-admitted. At runtime, a lightweight traverser matches the base model's intent to child macros, executes one macro at a time under global and node-local gating, and when stalled performs risk-aware shortest-path recovery to a feasible success leaf; the visited path forms a compact spine memory that replaces transcript replay. Evaluated in a unified OpenHands sandbox on 15+ software, web, reasoning, and safety/security benchmarks, GBT improves success while driving violations toward zero and reducing cost. On SWE-bench Verified (Protocol A, 500 issues), GBT-SE raises success from 34.6% to 73.6%, reduces violations from 2.8% to 0.2%, and cuts token/character usage from 208k/820k to 126k/490k; with the same distilled tree, 8B executors more than double success on SWE-bench Verified (14.0%58.8%) and WebArena (9.1%37.3%).

cs.LG↗

arXiv2025

AutoCode: LLMs as Problem Setters for Competitive Programming

Writing competitive programming problems is exacting. Authors must: set constraints, input distributions, and edge cases that rule out shortcuts; target specific algorithms (e.g., max-flow, dynamic programming, data structures); and calibrate complexity beyond the reach of most competitors. We argue that this makes for an ideal test of general large language model capabilities and study whether they can do this reliably. We introduce AutoCode, which uses multiple rounds of validation to yield competition-grade problem statements and test cases. On held-out problems, AutoCode test suites approach 99% consistency with official judgments, a significant improvement over current state-of-the-art methods like HardTests, which achieve less than 81%. Furthermore, starting with a random seed problem, AutoCode can create novel variants with reference and brute-force solutions. By cross-verifying these generated solutions against test cases, we can further filter out malformed problems. Our system ensures high correctness, as verified by human experts. AutoCode successfully produces novel problems judged by Grandmaster-level (top 0.3%) competitive programmers to be of contest quality.

cs.SE↗

arXiv2022

Quantum Melting of Spin-1 Dimer Solid Induced by Inter-chain Couplings

Dimerized valence bond solids appear naturally in spin-1/2 systems on bipartite lattices, with the geometric frustrations playing a key role both in their stability and the eventual `melting' due to quantum fluctuations. Here, we ask the question of the stability of such dimerized solids in spin-1 systems, taking the anisotropic square lattice with bilinear and biquadratic spin-spin interactions as a paradigmatic model. The lattice can be viewed as a set of coupled spin-1 chains, which in the limit of vanishing inter-chain coupling are known to possess a stable dimer phase. We study this model using the density matrix renormalization group (DMRG) and infinite projected entangled-pair states (iPEPS) techniques, supplemented by the analytical mean-field and linear flavor wave theory calculations. While the latter predicts the dimer phase to remain stable up to a reasonably large interchain-to-intrachain coupling ratio $r \lesssim 0.6$, the DMRG and iPEPS find that the dimer solid melts for much weaker interchain coupling not exceeding $r\lesssim 0.15$. We find the transition into a magnetically ordered state to be first order, manifested by a hysteresis and order parameter jump, precluding the deconfined quantum critical scenario. The apparent lack of stability of dimerized phases in 2D spin-1 systems is indicative of strong quantum fluctuations that melt the dimer solid.

cond-mat.str-el↗