Ctx2Skill lets language models autonomously evolve context-specific skills via multi-agent self-play, improving performance on context learning tasks without human supervision.
hub
Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains
24 Pith papers cite this work. Polarity classification is still indexing.
24
Pith papers citing it
abstract
Reinforcement Learning with Verifiable Rewards (RLVR) has proven effective for complex reasoning tasks with clear correctness signals such as math and coding. However, extending it to real-world reasoning tasks is challenging, as evaluation depends on nuanced, multi-criteria judgments rather than binary correctness. Instance-specific rubrics have recently been used in evaluation benchmarks to capture such judgments, but their potential as reward signals for on-policy post-training remains underexplored. We introduce $\textbf{Rubrics as Rewards}$ (RaR), an on-policy reinforcement learning method that extends RLVR beyond verifiable domains by using rubric-based feedback. Across both medical and science domains, we evaluate multiple strategies for aggregating rubric feedback into rewards. The best RaR variant achieves relative improvements of up to $31\%$ on HealthBench and $7\%$ on GPQA-Diamond over popular LLM-as-judge baselines that rely on direct Likert-based rewards. These results demonstrate that RaR-trained policies adapt well to diverse evaluation formats, performing strongly on both rubric-based and multiple-choice tasks. Moreover, we find that using rubrics as structured reward signals yields better alignment for smaller judges and reduces performance variance across judge scales.
hub tools
citation-role summary
background 1
citation-polarity summary
years
2026 24roles
background 1polarities
background 1representative citing papers
Think-with-Rubrics has LLMs generate rubrics internally before responding, outperforming external rubric-as-reward baselines by 3.87 points on average across benchmarks.
Rubric-based on-policy distillation allows training student models using only teacher responses by generating scoring rubrics from contrasts and using them for on-policy optimization, achieving superior performance and up to 10x better sample efficiency than logit-based approaches.
rDPO uses offline-built rubrics to generate on-policy preference data for DPO, raising benchmark scores in visual tasks over outcome-based filtering and style baselines.
Rubric-based LLM judges show self-preference bias, incorrectly marking their own failed outputs as satisfied up to 50% more often on verifiable benchmarks and skewing scores by 10 points on subjective ones.
DAgger-style training with turn-level policy interpolation raises 4B and 8B LLM agents to 27.3% and 29.8% on SWE-bench Verified, beating several larger published systems.
Rubric-based RL verifiers can be gamed via partial criterion satisfaction and implicit-to-explicit tricks, yielding proxy gains that do not improve quality under rubric-free judges; stronger verifiers reduce but do not eliminate the mismatch.
SAGE trains a rubric-based verifier and an RL-optimized generator on seed human data to scalably augment LLM knowledge benchmarks, matching human-annotated quality on HellaSwag at lower cost and generalizing to MMLU.
RubricRefine raises average tool-use reliability to 0.86 on M3ToolEval across seven models by scoring candidate code against generated contract rubrics before execution, beating prior inference-time methods at 2.6X lower latency.
CLR-voyance reformulates inpatient reasoning as POMDP with clinician-validated outcome rubrics, yielding an 8B model that outperforms larger frontier models on the authors' new benchmark.
DeltaRubric decomposes multimodal preference evaluation into self-generated planning and verification steps within a single model, producing large accuracy improvements on VL-RewardBench via multi-role reinforcement learning.
Rubric-grounded RL with LLM judges on document-derived criteria raises Llama-3.1-8B normalized reward to 71.7% on held-out rubrics and improves performance on GSM8K, MATH, and GPQA benchmarks.
SHARP is a neuro-symbolic method that evolves bounded, auditable rule rubrics for LLM trading agents via cross-sample attribution and walk-forward validation, raising compact-model performance by 10-20 percentage points across equity sectors.
RVPO penalizes variance across multiple reward signals during RLHF advantage aggregation, using a LogSumExp operator as a smooth variance penalty to reduce constraint neglect in LLM alignment.
Edit-R1 trains a CoT-based reasoning reward model with GCPO and uses it to boost image editing performance over VLMs and models like FLUX.1-kontext via GRPO.
POP bootstraps post-training signals for open-ended LLM tasks by synthesizing rubrics during self-play on pretraining corpus, yielding performance gains on Qwen-2.5-7B across healthcare QA, creative writing, and instruction following.
C2 synthesizes contrastive helpful/misleading rubric pairs from binary preferences to train cooperative generators and critical verifiers, yielding up to 6.5-point gains on RM-Bench and enabling smaller models to match larger rubric-augmented ones.
ReflectRM improves generative reward models by adding self-reflection on analysis quality within a unified training setup for response and analysis preferences, yielding accuracy gains and reduced positional bias on benchmarks.
Systematic false positives in verifiers can cause RLVR training to reach suboptimal plateaus or collapse, with outcomes driven by error patterns rather than overall error rate.
Fine-tuned simulators grounded in real human data produce LLM assistants that win more often against real users than those trained against role-playing simulators.
SCPRM adds prefix conditioning and schema distance to process reward models so that Monte Carlo Tree Search can explore knowledge-graph reasoning paths with both cumulative and future guidance, yielding a 1.18% average Hits@k gain on medical, legal, and CWQ tasks.
LegalDrill uses diagnosis-driven synthesis and self-reflective verification to create high-quality training data that improves small language models' legal reasoning without expert annotations.
PubSwap uses a small public dataset for selective off-policy response swapping in federated RLVR to improve coordination and performance over standard baselines on math and medical reasoning tasks.
SPARD dynamically tunes multi-objective reward weights and data importance in LLM reinforcement learning alignment using a self-paced curriculum driven by reward dynamics and data utility.
citing papers explorer
-
From Context to Skills: Can Language Models Learn from Context Skillfully?
Ctx2Skill lets language models autonomously evolve context-specific skills via multi-agent self-play, improving performance on context learning tasks without human supervision.
-
Think-with-Rubrics: From External Evaluator to Internal Reasoning Guidance
Think-with-Rubrics has LLMs generate rubrics internally before responding, outperforming external rubric-as-reward baselines by 3.87 points on average across benchmarks.
-
Rubric-based On-policy Distillation
Rubric-based on-policy distillation allows training student models using only teacher responses by generating scoring rubrics from contrasts and using them for on-policy optimization, achieving superior performance and up to 10x better sample efficiency than logit-based approaches.
-
Visual Preference Optimization with Rubric Rewards
rDPO uses offline-built rubrics to generate on-policy preference data for DPO, raising benchmark scores in visual tasks over outcome-based filtering and style baselines.
-
Self-Preference Bias in Rubric-Based Evaluation of Large Language Models
Rubric-based LLM judges show self-preference bias, incorrectly marking their own failed outputs as satisfied up to 50% more often on verifiable benchmarks and skewing scores by 10 points on subjective ones.
-
Revisiting DAgger in the Era of LLM-Agents
DAgger-style training with turn-level policy interpolation raises 4B and 8B LLM agents to 27.3% and 29.8% on SWE-bench Verified, beating several larger published systems.
-
Reward Hacking in Rubric-Based Reinforcement Learning
Rubric-based RL verifiers can be gamed via partial criterion satisfaction and implicit-to-explicit tricks, yielding proxy gains that do not improve quality under rubric-free judges; stronger verifiers reduce but do not eliminate the mismatch.
-
SAGE: Scalable Automated Robustness Augmentation for LLM Knowledge Evaluation
SAGE trains a rubric-based verifier and an RL-optimized generator on seed human data to scalably augment LLM knowledge benchmarks, matching human-annotated quality on HellaSwag at lower cost and generalizing to MMLU.
-
RubricRefine: Improving Tool-Use Agent Reliability with Training-Free Pre-Execution Refinement
RubricRefine raises average tool-use reliability to 0.86 on M3ToolEval across seven models by scoring candidate code against generated contract rubrics before execution, beating prior inference-time methods at 2.6X lower latency.
-
CLR-voyance: Reinforcing Open-Ended Reasoning for Inpatient Clinical Decision Support with Outcome-Aware Rubrics
CLR-voyance reformulates inpatient reasoning as POMDP with clinician-validated outcome rubrics, yielding an 8B model that outperforms larger frontier models on the authors' new benchmark.
-
DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification
DeltaRubric decomposes multimodal preference evaluation into self-generated planning and verification steps within a single model, producing large accuracy improvements on VL-RewardBench via multi-role reinforcement learning.
-
Rubric-Grounded RL: Structured Judge Rewards for Generalizable Reasoning
Rubric-grounded RL with LLM judges on document-derived criteria raises Llama-3.1-8B normalized reward to 71.7% on held-out rubrics and improves performance on GSM8K, MATH, and GPQA benchmarks.
-
SHARP: A Self-Evolving Human-Auditable Rubric Policy for Financial Trading Agents
SHARP is a neuro-symbolic method that evolves bounded, auditable rule rubrics for LLM trading agents via cross-sample attribution and walk-forward validation, raising compact-model performance by 10-20 percentage points across equity sectors.
-
RVPO: Risk-Sensitive Alignment via Variance Regularization
RVPO penalizes variance across multiple reward signals during RLHF advantage aggregation, using a LogSumExp operator as a smooth variance penalty to reduce constraint neglect in LLM alignment.
-
Leveraging Verifier-Based Reinforcement Learning in Image Editing
Edit-R1 trains a CoT-based reasoning reward model with GCPO and uses it to boost image editing performance over VLMs and models like FLUX.1-kontext via GRPO.
-
Bootstrapping Post-training Signals for Open-ended Tasks via Rubric-based Self-play on Pre-training Text
POP bootstraps post-training signals for open-ended LLM tasks by synthesizing rubrics during self-play on pretraining corpus, yielding performance gains on Qwen-2.5-7B across healthcare QA, creative writing, and instruction following.
-
C2: Scalable Rubric-Augmented Reward Modeling from Binary Preferences
C2 synthesizes contrastive helpful/misleading rubric pairs from binary preferences to train cooperative generators and critical verifiers, yielding up to 6.5-point gains on RM-Bench and enabling smaller models to match larger rubric-augmented ones.
-
ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework
ReflectRM improves generative reward models by adding self-reflection on analysis quality within a unified training setup for response and analysis preferences, yielding accuracy gains and reduced positional bias on benchmarks.
-
Delay, Plateau, or Collapse: Evaluating the Impact of Systematic Verification Error on RLVR
Systematic false positives in verifiers can cause RLVR training to reach suboptimal plateaus or collapse, with outcomes driven by error patterns rather than overall error rate.
-
Quantifying the Utility of User Simulators for Building Collaborative LLM Assistants
Fine-tuned simulators grounded in real human data produce LLM assistants that win more often against real users than those trained against role-playing simulators.
-
SCPRM: A Schema-aware Cumulative Process Reward Model for Knowledge Graph Question Answering
SCPRM adds prefix conditioning and schema distance to process reward models so that Monte Carlo Tree Search can explore knowledge-graph reasoning paths with both cumulative and future guidance, yielding a 1.18% average Hits@k gain on medical, legal, and CWQ tasks.
-
LegalDrill: Diagnosis-Driven Synthesis for Legal Reasoning in Small Language Models
LegalDrill uses diagnosis-driven synthesis and self-reflective verification to create high-quality training data that improves small language models' legal reasoning without expert annotations.
-
PubSwap: Public-Data Off-Policy Coordination for Federated RLVR
PubSwap uses a small public dataset for selective off-policy response swapping in federated RLVR to improve coordination and performance over standard baselines on math and medical reasoning tasks.
-
SPARD: Self-Paced Curriculum for RL Alignment via Integrating Reward Dynamics and Data Utility
SPARD dynamically tunes multi-objective reward weights and data importance in LLM reinforcement learning alignment using a self-paced curriculum driven by reward dynamics and data utility.