TAC is a bandit curriculum for multi-domain RLVR that prioritizes domains whose gradient updates align with and benefit other domains, yielding up to 2.8-point macro accuracy gains over learnability-only baselines on Qwen3-1.7B and Llama3.2-3B.
hub
Leetcodedataset: A temporal dataset for robust evaluation and efficient training of code llms
16 Pith papers cite this work. Polarity classification is still indexing.
16
Pith papers citing it
hub tools
citation-role summary
dataset 1
citation-polarity summary
roles
dataset 1polarities
use dataset 1representative citing papers
Monitors trained on prompt-elicited reward-hacking trajectories fail to generalize to hacking behaviors that arise naturally during RL training of code models, whereas trajectories curated by Trace-and-Amplify transfer better and to unseen hack types.
CodeSpecBench shows LLMs achieve at most 20.2% pass rate on repository-level executable behavioral specification generation, revealing that strong code generation does not imply deep semantic understanding.
PRepair mitigates LLM over-editing in code repair via Self-Breaking bug injection and EA-GRPO training, improving precision by up to 31.4% on the fix₁@1 metric.
Think-Anywhere lets LLMs invoke on-demand reasoning at any token during code generation via cold-start imitation followed by outcome-based RL, reaching state-of-the-art results on LeetCode, LiveCodeBench, HumanEval, and MBPP.
IFCodeEvolve synthesizes coding data via actor-schema co-evolution with MCTS, boosting a 32B model's performance to match proprietary SOTA on instruction following.
CodeRL+ integrates variable-level execution trajectory inference into RLVR training to align textual code representations with execution semantics, delivering 4.6% relative pass@1 gains and generalization to code-reasoning and test-output tasks.
Reasoning-tuned LLMs align with human comprehension failure patterns under code obfuscation using the Block Model, unlike instruction-tuned variants.
MONA integrates Nesterov acceleration into Muon's orthogonalization framework, reporting better convergence than Muon and AdamW on MoE models up to 68B parameters trained on 1T tokens and SOTA fine-tuning results.
CodeThinker improves LLM code reasoning via consistency-based RL with stepwise training data, dynamic beam sampling, and consistency rewards, reaching SOTA on benchmarks with 4.3% gains on Qwen2.5-Coder-7B.
CIPO jointly optimizes standard RLVR rewards with correction samples derived from the model's own failed attempts, yielding better reasoning and self-correction on math and code benchmarks.
Hybrid-LoRA selectively full fine-tunes modules with high sensitivity to low-rank adaptation using a novel score and applies LoRA elsewhere, matching full fine-tuning at 10% budget and outperforming PEFT baselines by up to 5.65%.
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.
PerfOrch is a four-agent multi-LLM system that uses offline profiling to build language-and-category rankings for routing tasks, achieving 97.19% and 95.83% pass@1 on HumanEval-X and EffiBench-X with generalization across benchmarks.
A step-wise reweighting of on-policy distillation, based on per-step student-teacher divergence, improves tool-integrated reasoning in 0.6B and 1.7B language models.
Skywork-OR1 uses RL on distilled CoT models to lift math and coding benchmark accuracy by 13-15 points while open-sourcing everything.
citing papers explorer
-
Transferability for General Reasoning: An Automated Curriculum for Multi-Domain RLVR
TAC is a bandit curriculum for multi-domain RLVR that prioritizes domains whose gradient updates align with and benefit other domains, yielding up to 2.8-point macro accuracy gains over learnability-only baselines on Qwen3-1.7B and Llama3.2-3B.
-
Do Prompt-Elicited Trajectories Reflect Training-Time Reward Hacking? A Systematic Study on Monitoring Training-Time Reward Hacking in Code Generation
Monitors trained on prompt-elicited reward-hacking trajectories fail to generalize to hacking behaviors that arise naturally during RL training of code models, whereas trajectories curated by Trace-and-Amplify transfer better and to unseen hack types.
-
CodeSpecBench: Benchmarking LLMs for Executable Behavioral Specification Generation
CodeSpecBench shows LLMs achieve at most 20.2% pass rate on repository-level executable behavioral specification generation, revealing that strong code generation does not imply deep semantic understanding.
-
QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization
PRepair mitigates LLM over-editing in code repair via Self-Breaking bug injection and EA-GRPO training, improving precision by up to 31.4% on the fix₁@1 metric.
-
Think Anywhere in Code Generation
Think-Anywhere lets LLMs invoke on-demand reasoning at any token during code generation via cold-start imitation followed by outcome-based RL, reaching state-of-the-art results on LeetCode, LiveCodeBench, HumanEval, and MBPP.
-
Steerable Instruction Following Coding Data Synthesis with Actor-Parametric Schema Co-Evolution
IFCodeEvolve synthesizes coding data via actor-schema co-evolution with MCTS, boosting a 32B model's performance to match proprietary SOTA on instruction following.
-
CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignment
CodeRL+ integrates variable-level execution trajectory inference into RLVR training to align textual code representations with execution semantics, delivering 4.6% relative pass@1 gains and generalization to code-reasoning and test-output tasks.
-
Do Machines Struggle Where Humans Do? LLM and Human Comprehension of Obfuscated Code
Reasoning-tuned LLMs align with human comprehension failure patterns under code obfuscation using the Block Model, unlike instruction-tuned variants.
-
MONA: Muon Optimizer with Nesterov Acceleration for Scalable Language Model Training
MONA integrates Nesterov acceleration into Muon's orthogonalization framework, reporting better convergence than Muon and AdamW on MoE models up to 68B parameters trained on 1T tokens and SOTA fine-tuning results.
-
Enhancing the Code Reasoning Capabilities of LLMs via Consistency-based Reinforcement Learning
CodeThinker improves LLM code reasoning via consistency-based RL with stepwise training data, dynamic beam sampling, and consistency rewards, reaching SOTA on benchmarks with 4.3% gains on Qwen2.5-Coder-7B.
-
Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards
CIPO jointly optimizes standard RLVR rewards with correction samples derived from the model's own failed attempts, yielding better reasoning and self-correction on math and code benchmarks.
-
Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training
Hybrid-LoRA selectively full fine-tunes modules with high sensitivity to low-rank adaptation using a novel score and applies LoRA elsewhere, matching full fine-tuning at 10% budget and outperforming PEFT baselines by up to 5.65%.
-
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.
-
Multi-LLM Orchestration for High-Quality Code Generation: Exploiting Complementary Model Strengths
PerfOrch is a four-agent multi-LLM system that uses offline profiling to build language-and-category rankings for routing tasks, achieving 97.19% and 95.83% pass@1 on HumanEval-X and EffiBench-X with generalization across benchmarks.
-
SOD: Step-wise On-policy Distillation for Small Language Model Agents
A step-wise reweighting of on-policy distillation, based on per-step student-teacher divergence, improves tool-integrated reasoning in 0.6B and 1.7B language models.
-
Skywork Open Reasoner 1 Technical Report
Skywork-OR1 uses RL on distilled CoT models to lift math and coding benchmark accuracy by 13-15 points while open-sourcing everything.