HM-Bench is the first benchmark for MLLMs on hyperspectral images, showing models struggle with complex spatial-spectral reasoning and perform better with visual PCA images than textual reports.
super hub
Qwen2.5-Coder Technical Report
127 Pith papers cite this work. Polarity classification is still indexing.
127
Pith papers citing it
abstract
In this report, we introduce the Qwen2.5-Coder series, a significant upgrade from its predecessor, CodeQwen1.5. This series includes six models: Qwen2.5-Coder-(0.5B/1.5B/3B/7B/14B/32B). As a code-specific model, Qwen2.5-Coder is built upon the Qwen2.5 architecture and continues pretrained on a vast corpus of over 5.5 trillion tokens. Through meticulous data cleaning, scalable synthetic data generation, and balanced data mixing, Qwen2.5-Coder demonstrates impressive code generation capabilities while retaining general and math skills. These models have been evaluated on a wide range of code-related tasks, achieving state-of-the-art (SOTA) performance across more than 10 benchmarks, including code generation, completion, reasoning, and repair, consistently outperforming larger models of the same model size. We believe that the release of the Qwen2.5-Coder series will advance research in code intelligence and, with its permissive licensing, support wider adoption by developers in real-world applications.
hub tools
claims ledger
- abstract In this report, we introduce the Qwen2.5-Coder series, a significant upgrade from its predecessor, CodeQwen1.5. This series includes six models: Qwen2.5-Coder-(0.5B/1.5B/3B/7B/14B/32B). As a code-specific model, Qwen2.5-Coder is built upon the Qwen2.5 architecture and continues pretrained on a vast corpus of over 5.5 trillion tokens. Through meticulous data cleaning, scalable synthetic data generation, and balanced data mixing, Qwen2.5-Coder demonstrates impressive code generation capabilities while retaining general and math skills. These models have been evaluated on a wide range of code-rel
authors
co-cited works
representative citing papers
The first SoK on LLM-based AutoPT frameworks provides a six-dimension taxonomy of agent designs and a unified empirical benchmark evaluating 15 frameworks via over 10 billion tokens and 1,500 manually reviewed logs.
The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.
HLS-Seek replaces full-synthesis RL with a comparative proxy reward model plus uncertainty-triggered real checks, yielding higher correctness and better QoR than larger models at 8.5x lower training cost.
Reward-Weighted On-Policy Distillation with an open property-equivalence verifier produces a 7B model that surpasses prior SOTA on NL-to-SVA generation across pass@1/5/10 metrics.
A parallel multi-turn medical dialogue dataset spanning English and nine Indic languages is created from synthetic consultations to enable personalized AI healthcare interactions.
PyRAG turns multi-hop reasoning into executable Python code over retrieval tools for explicit, verifiable step-by-step RAG.
VLMs show a resolution illusion on UHR Earth observation imagery where higher resolution does not improve micro-target perception; UHR-Micro benchmark and MAP-Agent address this via evidence-centered active inspection.
dGRPO merges outcome-based policy optimization with dense teacher guidance from on-policy distillation, yielding more stable long-context reasoning on the new LongBlocks synthetic dataset.
DGAO uses reinforcement learning to optimize LLMs for both accuracy and order stability by balancing intra-group accuracy advantages and inter-group stability advantages.
StepCodeReasoner aligns code reasoning with verifiable stepwise execution traces via print anchors and bi-level GRPO reinforcement learning, reaching SOTA results on CRUXEval (91.1%) and LiveCodeBench (86.5%) for a 7B model.
PlantMarkerBench supplies 5,550 literature sentences annotated for plant marker gene evidence validity and type across Arabidopsis, maize, rice and tomato, showing frontier LLMs handle direct expression evidence but struggle with functional, indirect and weak-support cases.
Malicious Skills induce coding agents to hallucinate and import attacker-controlled packages at high rates while evading detection.
Mean-pooled cosine similarity grows with sequence length in anisotropic transformer embeddings independent of content, while CKA shows far less length dependence across code, translation, and vision tasks.
Weblica scales RL training for visual web agents by building thousands of reproducible environments through HTTP caching for stable replays and LLM synthesis from real sites, yielding an 8B model that beats similar open baselines on navigation benchmarks.
Fine-tuned 7B LLMs generating unified diffs for neural architecture refinement achieve 66-75% valid rates and 64-66% mean first-epoch accuracy, outperforming full-generation baselines by large margins while cutting output length by 75-85%.
Stream-R1 improves distillation of autoregressive streaming video diffusion models by adaptively weighting supervision with a reward model at both rollout and per-pixel levels.
ARISE adds a data-flow-augmented repository graph and three-tier tool API to LLM agents, raising Function Recall@1 by 17 points, Line Recall@1 by 15 points, and Pass@1 repair rate to 22% on SWE-bench Lite.
LiveFMBench shows that direct LLM prompting for C program formal specs overestimates accuracy by ~20% due to unfaithful behaviors like deceiving provers, while agentic workflows help under low sampling but overall performance remains far below human-authored specs.
Structurally rich task descriptions make LLMs robust to prompt under-specification, and under-specification can enhance code correctness by disrupting misleading lexical or structural cues.
Orchid benchmark shows requirement ambiguity degrades LLM code generation performance across all models, with advanced models hit hardest, and LLMs rarely detect or resolve the ambiguity themselves.
A cascaded large-small model system generates edit sketches with the large model and applies them with the small model to make code editing both accurate and token-efficient.
LLM planners for robots often produce dangerous plans even when planning succeeds, with safety awareness staying flat as model scale improves planning ability.
TriMix dynamically fuses logits from three model sources to outperform baselines and Proxy Tuning on eight low-resource languages across four model families.
citing papers explorer
-
POETS: Uncertainty-Aware LLM Optimization via Compute-Efficient Policy Ensembles
POETS uses compute-efficient LLM policy ensembles to implicitly perform KL-regularized Thompson sampling, delivering O(sqrt(T gamma_T)) regret bounds and state-of-the-art sample efficiency in scientific discovery tasks such as protein search and quantum circuit design.
-
Mage: Multi-Axis Evaluation of LLM-Generated Executable Game Scenes Beyond Compile-Pass Rate
Mage shows compile-pass rate is anti-correlated with functional correctness in LLM game scene generation; direct NL-to-C# yields 43% runtime but F1~0.12 structure, while IR conditioning recovers structure (F1 up to 1.0) but halves runtime, with granularity levels statistically equivalent.
-
PaT: Planning-after-Trial for Efficient Test-Time Code Generation
PaT defers planning until after failed trials in LLM code generation, enabling heterogeneous cheap-plus-powerful model setups that match large-model performance at roughly 69% lower cost.
-
Schedule-and-Calibrate: Utility-Guided Multi-Task Reinforcement Learning for Code LLMs
ASTOR improves a single code LLM across four tasks by 9.0-9.5% over the best specialist and 7.5-12.8% over prior multi-task RL baselines via utility-driven data scheduling and adaptive KL regularization.
-
Requests of a Feather Must Flock Together: Batch Size vs. Prefix Homogeneity in LLM Inference
Feather uses reinforcement learning and a Chunked Hash Tree to balance batch size against prefix homogeneity in LLM inference, delivering 2-10x higher throughput than existing schedulers.
-
On the Role of Language Representations in Auto-Bidding: Findings and Implications
SemBid injects LLM-encoded Task, History, and Strategy semantics as tokens into offline bidding trajectories and uses self-attention to outperform numerical-only baselines in performance, constraint satisfaction, and robustness.
-
Conceptors for Semantic Steering
Conceptors as soft projection matrices from bipolar activations offer a multidimensional, compositional, and geometrically principled method for semantic steering in LLMs that outperforms single-vector baselines in multi-dimensional subspaces.
-
SynConfRoute: Syntax-Aware Routing for Efficient Code Completion with Small CodeLLMs
SynConfRoute routes code completions using syntax validation and token confidence, improving pass@1 by up to 31% on hard tasks and reducing accelerator usage by 58% versus always using the largest model.
-
Improving LLM Code Generation via Requirement-Aware Curriculum Reinforcement Learning
REC RL improves LLM code generation by automatically assessing and optimizing requirement difficulty with adaptive curriculum sampling, yielding 1.23-5.62% Pass@1 gains over baselines.
-
Defective Task Descriptions in LLM-Based Code Generation: Detection and Analysis
SpecValidator detects lexical vagueness, under-specification, and syntax-formatting defects in LLM code-generation prompts with F1 0.804, outperforming GPT-5-mini and Claude Sonnet 4, and shows that under-specification is the most damaging defect type while richer benchmarks are more resilient.
-
Leveraging LLMs for Multi-File DSL Code Generation: An Industrial Case Study
Fine-tuning 7B code LLMs on a custom multi-file DSL dataset achieves structural fidelity of 1.00, high exact-match accuracy, and practical utility validated by expert survey and execution checks.
-
MEMCoder: Multi-dimensional Evolving Memory for Private-Library-Oriented Code Generation
MEMCoder boosts LLM code generation for private libraries by 16.31% pass@1 via a multi-dimensional evolving memory that distills usage guidelines from execution feedback and combines them with static docs.
-
Exploring High-Order Self-Similarity for Video Understanding
The MOSS module learns and combines multi-order space-time self-similarity features to enhance temporal dynamics modeling in videos across action recognition, VQA, and robotic tasks.
-
WebGen-R1: Incentivizing Large Language Models to Generate Functional and Aesthetic Websites with Reinforcement Learning
WebGen-R1 uses end-to-end RL with scaffold-driven generation and cascaded rewards for structure, function, and aesthetics to transform a 7B model into a generator of deployable multi-page websites that rivals much larger models.
-
Hybrid Policy Distillation for LLMs
Hybrid Policy Distillation unifies existing knowledge distillation methods for LLMs into a reweighted log-likelihood objective and introduces a hybrid forward-reverse KL approach with mixed data sampling to improve stability, efficiency, and performance.
-
To Know is to Construct: Schema-Constrained Generation for Agent Memory
SCG-MEM reformulates agent memory access as schema-constrained generation within dynamic cognitive schemas, using assimilation and accommodation for updates plus an associative graph for reasoning, and outperforms retrieval baselines on the LoCoMo benchmark.
-
SAGE: Signal-Amplified Guided Embeddings for LLM-based Vulnerability Detection
SAGE uses sparse autoencoders to boost vulnerability signals in LLMs, raising internal SNR 12.7x and delivering up to 318% MCC gains on vulnerability detection benchmarks.
-
AutoPPA: Automated Circuit PPA Optimization via Contrastive Code-based Rule Library Learning
AutoPPA learns generalizable PPA optimization rules automatically via contrastive abstraction from diverse code pairs and applies them through adaptive search, outperforming manual methods and prior tools SymRTLO and RTLRewriter.
-
Co-Located Tests, Better AI Code: How Test Syntax Structure Affects Foundation Model Code Generation
Co-locating tests with implementation code yields substantially higher preservation and correctness in foundation-model-generated programs than separated test syntax.
-
PARM: Pipeline-Adapted Reward Model
PARM adapts reward models to multi-stage LLM pipelines via pipeline data and direct preference optimization, improving execution rate and solving accuracy on optimization benchmarks and showing transfer to GSM8K.
-
LeGo-Code: Can Modular Curriculum Learning Advance Complex Code Generation? Insights from Text-to-SQL
Modular curriculum learning with tier-specific adapters outperforms standard fine-tuning on complex Text-to-SQL queries in Spider and BIRD benchmarks by avoiding catastrophic forgetting.
-
Modular Representation Compression: Adapting LLMs for Efficient and Effective Recommendations
LLMs exhibit mid-layer representation advantage for recommendations; MARC compresses representations modularly to reduce costs while improving performance, as shown in a large-scale online advertising deployment.
-
Program Structure-aware Language Models: Targeted Software Testing beyond Textual Semantics
GLMTest integrates code property graphs and GNNs with LLMs to steer test case generation toward targeted branches, raising branch accuracy from 27.4% to 50.2% on the TestGenEval benchmark.
-
MM-WebAgent: A Hierarchical Multimodal Web Agent for Webpage Generation
MM-WebAgent is a hierarchical multimodal agent that coordinates AIGC tools through planning and iterative self-reflection to generate coherent, visually consistent webpages and outperforms baselines on a new benchmark.
-
DPC: Training-Free Text-to-SQL Candidate Selection via Dual-Paradigm Consistency
DPC selects correct text-to-SQL outputs by enforcing execution consistency between SQL and Python on an adversarially constructed minimal distinguishing database.
-
On the Effectiveness of Context Compression for Repository-Level Tasks: An Empirical Investigation
Continuous latent-vector compression improves BLEU scores on repository-level code tasks by up to 28.3% at 4x compression while cutting inference latency.
-
TOPCELL: Topology Optimization of Standard Cell via LLMs
TOPCELL reformulates standard cell topology optimization as an LLM generative task with GRPO fine-tuning, outperforming base models and matching exhaustive solvers with 85.91x speedup in 2nm/7nm industrial flows.
-
Generating Effective CoT Traces for Mitigating Causal Hallucination
A pipeline generates CoT traces that reduce causal hallucination in small LLMs on event causality tasks, paired with a new Causal Hallucination Rate metric that guides and validates the process.
-
TEMPLATEFUZZ: Fine-Grained Chat Template Fuzzing for Jailbreaking and Red Teaming LLMs
TEMPLATEFUZZ mutates chat templates with element-level rules and heuristic search to reach 98.2% average jailbreak success rate on twelve open-source LLMs while degrading accuracy by only 1.1%.
-
Verify Before You Fix: Agentic Execution Grounding for Trustworthy Cross-Language Code Analysis
A framework combining universal AST normalization, hybrid graph-LLM embeddings, and strict execution-grounded validation achieves 89-92% intra-language accuracy and 74-80% cross-language F1 while resolving 70% of vulnerabilities at 12% failure rate.
-
Strix: Re-thinking NPU Reliability from a System Perspective
Strix delivers sub-microsecond fault localisation, detection, and correction on NPUs with 1.04x slowdown and minimal hardware cost by system-level re-partitioning and targeted safeguards.
-
When LLMs Lag Behind: Knowledge Conflicts from Evolving APIs in Code Generation
LLMs produce executable code only 42.55% of the time under API evolution without full documentation, improving to 66.36% with structured docs and by 11% more with reasoning strategies, yet outdated patterns persist.
-
DeepGuard: Secure Code Generation via Multi-Layer Semantic Aggregation
DeepGuard aggregates multi-layer representations in code LLMs to raise the secure-and-correct generation rate by 11.9% on average over baselines like SVEN while preserving correctness and generalizing to new vulnerability types.
-
Distributed Multi-Layer Editing for Rule-Level Knowledge in Large Language Models
Rule knowledge in LLMs is localized by form across layers; a distributed multi-layer editing method improves instance portability by 13.91 and rule understanding by 50.19 percentage points over baselines on multiple models.
-
Uni-ViGU: Towards Unified Video Generation and Understanding via A Diffusion-Based Video Generator
Uni-ViGU unifies video generation and understanding by extending a diffusion video generator with unified continuous-discrete flow matching, modality-driven MoE layers, and bidirectional training stages that repurpose generative knowledge for discriminative tasks.
-
ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?
ZeroCoder co-evolves coder and tester LLMs via self-generated code-test execution feedback to improve code generation up to 21.6% without ground-truth supervision.
-
LLM4CodeRE: Generative AI for Code Decompilation Analysis and Reverse Engineering
LLM4CodeRE adapts LLMs with multi-adapter and seq2seq fine-tuning for accurate assembly-to-source decompilation and reverse translation in code reverse engineering.
-
PROMISE: Proof Automation as Structural Imitation of Human Reasoning
PROMISE reframes automated proof generation as stateful search over structural embeddings of proof states, outperforming prior LLM-based systems by up to 26 points on the seL4 benchmark.
-
InCoder-32B-Thinking: Industrial Code World Model for Thinking
InCoder-32B-Thinking uses error-feedback synthesized thinking traces and a code world model to reach top open-source scores on general and industrial code benchmarks including 81.3% on LiveCodeBench and 84.0% on CAD-Coder.
-
TestDecision: Sequential Test Suite Generation via Greedy Optimization and Reinforcement Learning
By proving test suite coverage is monotone submodular and training LLMs with RL to maximize marginal gains, TestDecision improves branch coverage 38-52% and bug detection up to 95% over base models on ULT and LiveCodeBench.
-
From SWE-ZERO to SWE-HERO: Execution-free to Execution-based Fine-tuning for Software Engineering Agents
A two-stage SFT pipeline distills execution-free then execution-based trajectories from a 480B model into smaller Qwen2.5-Coder agents, yielding 62.2% resolution on SWE-bench Verified and 44.1% zero-shot on the multilingual version.
-
Memory in the Age of AI Agents
The paper maps agent memory research via three forms (token-level, parametric, latent), three functions (factual, experiential, working), and dynamics of formation/evolution/retrieval, plus benchmarks and future directions.
-
Visual-RFT: Visual Reinforcement Fine-Tuning
Visual-RFT applies reinforcement learning with verifiable perception rewards to improve large vision-language models on fine-grained classification, few-shot detection, and grounding tasks.
-
Search-o1: Agentic Search-Enhanced Large Reasoning Models
Search-o1 integrates agentic retrieval-augmented generation and a Reason-in-Documents module into large reasoning models to dynamically supply missing knowledge and improve performance on complex science, math, coding, and QA tasks.
-
Scaling Retrieval-Augmented Reasoning with Parallel Search and Explicit Merging
MultiSearch uses parallel multi-query retrieval plus explicit merging inside a reinforcement-learning loop to improve retrieval-augmented reasoning, outperforming baselines on seven QA benchmarks.
-
Interactive Critique-Revision Training for Reliable Structured LLM Generation
DPA-GRPO trains a generator-verifier pair via group-relative policy optimization on paired counterfactual actions, improving structured output accuracy on TaxCalcBench over zero-shot and generator-only baselines.
-
Graph Representation Learning Augmented Model Manipulation on Federated Fine-Tuning of LLMs
Graph representation learning plus iterative augmented Lagrangian optimization creates stronger, harder-to-detect model manipulation attacks on federated LLM fine-tuning, cutting global accuracy by up to 26%.
-
HyperLens: Quantifying Cognitive Effort in LLMs with Fine-grained Confidence Trajectory
HyperLens reveals that deeper transformer layers magnify small confidence changes into fine-grained trajectories, allowing quantification of cognitive effort where complex tasks demand more and standard SFT can reduce it.
-
Attribution-Guided Continual Learning for Large Language Models
An attribution-based continual learning framework for LLMs modulates per-parameter gradients using task-specific importance scores to reduce forgetting of prior tasks.
-
How Does Chunking Affect Retrieval-Augmented Code Completion? A Controlled Empirical Study
Function-based chunking underperforms other strategies in RAG code completion by 3.57-5.64 points, with context length as the dominant factor.