RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization

• URL: http://arxiv.org/abs/2605.04539v3
• Date: Tue, 12 May 2026 05:46:48 GMT
• Title: RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization
• Abstract summary: We propose RLearner-LLM with Hybrid-DPO.<n>RLearner-LLM yields up to 6x NLI improvement over SFT.<n>On Gemma 4 E4B-it, Hybrid-DPO lifts NLI in four of five domains.
• Score: 5.0009109610311855
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Direct Preference Optimization (DPO), the efficient alternative to PPO-based RLHF, falls short on knowledge-intensive generation: standard preference signals from human annotators or LLM judges exhibit a systematic verbosity bias that rewards fluency over logical correctness. This blindspot leaves a logical alignment gap -- SFT models reach NLI entailment of only 0.05-0.22 despite producing fluent text. We propose RLearner-LLM with Hybrid-DPO: an automated preference pipeline that fuses a DeBERTa-v3 NLI signal with a verifier LLM score, removing human annotation while overcoming the "alignment tax" of single-signal optimization. Evaluated across five academic domains (Biology, Medicine, Law) with three base architectures (LLaMA-2-13B, Qwen3-8B, Gemma 4 E4B-it), RLearner-LLM yields up to 6x NLI improvement over SFT, with NLI gains in 11 of 15 cells and consistent answer-coverage gains. On Gemma 4 E4B-it (4.5B effective params), Hybrid-DPO lifts NLI in four of five domains (+11.9% to +2.4x) with faster inference across all five, scaling down to compact base models without losing the alignment-tax mitigation. Our Qwen3-8B RLearner-LLM wins 95% of pairwise comparisons against its own SFT baseline; GPT-4o-mini in turn wins 95% against our concise output -- alongside the 69% win the same judge gives a verbose SFT over our DPO model, this replicates verbosity bias on a frontier comparator and motivates logic-aware metrics (NLI, ACR) over LLM-as-a-judge for knowledge-intensive generation.

Related papers

• Triple-Bottom-Line Sustainability of Language Models for Edge AI: A Comparison Between SLMs and Quantized LLMs [0.0]
  Edge-AI model selection is commonly driven by one isolated metric - accuracy, latency, memory, energy, or safety.<n>Our work focuses on answering the question whether na- tively trained small language models (SLMs) or large language models (LLMs) compressed through post-training quantization offer the more sustainable edge- deployment trade-off.
  arXiv  Detail & Related papers  (2026-09-01T03:44:18Z)
• Towards Fast and Effective Long Video Understanding of Multimodal Large Language Models via Adaptive Quasi-Gaussian Sampling [71.62893745623526]
  Video frame selection is a problem of Quasi-Gaussian Sampling.<n>We propose an adaptive and training-free approach termed AdaQ.
  arXiv  Detail & Related papers  (2026-06-23T06:13:30Z)
• Reinforcement learning to improve large language model-based automated code compliance systems [1.0760732143259102]
  This paper introduces P4IR, a two-stage framework that uses supervised fine-tuning to instill domain knowledge in an LLM.<n>The framework achieves reductions of up to 23.8% and 38.6% in tree edit distance and token-level Levenshtein distance respectively.
  arXiv  Detail & Related papers  (2026-06-21T09:17:02Z)
• Towards Robust Training in NNGPT AutoML Pipeline: A Loss-Optimizer Pairing Selection Study [48.83701310501069]
  This paper investigates whether a single recipe is sufficient for heterogeneous architecture pools, or whether the optimal pairing varies across structurally diverse models.<n>We conduct a systematic empirical study of all $3 times 6 = 18$ combinations of six classifiers paired with three loss functions: Cross-Entropy (CEL), Negative Log-Likelihood (NLL), and the recently introduced genetically evolved NGL loss across the base models presented in LEMUR heterogeneous architecture pool on six image classification datasets.<n>Our results confirm that no single pairing is universally optimal. Cross-Entropy with Adam or AdamW is the most robust choice across
  arXiv  Detail & Related papers  (2026-06-18T20:51:42Z)
• SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning [28.61073266521457]
  A CPU-offloaded bug in DeepSpeed silently drops intermediate micro-batches during gradient accumulation.<n>A loss aggregation bug in OpenRLHF incorrectly weights per-mini-batch losses.<n>A truncated variant with just 50 RL steps outperforms mixed-policy methods on math benchmarks while using fewer FLOPs.
  arXiv  Detail & Related papers  (2026-04-26T14:53:48Z)
• JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency [56.46698214215968]
  JoyAI-LLM Flash is an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency.<n>JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline.
  arXiv  Detail & Related papers  (2026-04-03T13:52:38Z)
• REAL: Regression-Aware Reinforcement Learning for LLM-as-a-Judge [83.2858110368572]
  We propose textbfREAL (underlineREgression-underlineAware Reinforcement underlineLThought), a principled RL framework designed to optimize regression rewards.<n>We show that REAL consistently outperforms both regression-aware SFT baselines and standard RL methods.
  arXiv  Detail & Related papers  (2026-03-17T21:19:08Z)
• UniAPL: A Unified Adversarial Preference Learning Framework for Instruct-Following [12.924923059340395]
  We argue that post-training alignment is fundamentally a unified Preference Learning problem.<n>UniAPL implements a single-stage unified training objective, jointly learning from mixed batches of SFT and preference data.
  arXiv  Detail & Related papers  (2025-09-29T17:53:09Z)
• RIDE: Enhancing Large Language Model Alignment through Restyled In-Context Learning Demonstration Exemplars [57.6513924960128]
  Alignment tuning is crucial for ensuring large language models (LLMs) behave ethically and helpfully.<n>This paper proposes a low-cost, tuning-free method using in-context learning (ICL) to enhance LLM alignment.
  arXiv  Detail & Related papers  (2025-02-17T11:16:19Z)
• LLM2: Let Large Language Models Harness System 2 Reasoning [65.89293674479907]
  Large language models (LLMs) have exhibited impressive capabilities across a myriad of tasks, yet they occasionally yield undesirable outputs.<n>We introduce LLM2, a novel framework that combines an LLM with a process-based verifier.<n>LLMs2 is responsible for generating plausible candidates, while the verifier provides timely process-based feedback to distinguish desirable and undesirable outputs.
  arXiv  Detail & Related papers  (2024-12-29T06:32:36Z)
• Weighted-Reward Preference Optimization for Implicit Model Fusion [35.57286356489511]
  We propose an implicit fusion method, which leverages preference optimization between the source LLMs and the target LLM to transfer their capabilities effectively.<n>WRPO eliminates the need for vocabulary alignment and matrix fusion and can be efficiently scaled to accommodate various LLMs.<n>Experiments on the MT-Bench, AlpacaEval-2, and Arena-Hard benchmarks demonstrate that WRPO consistently outperforms existing knowledge fusion methods.
  arXiv  Detail & Related papers  (2024-12-04T10:15:12Z)
• Self-Play Preference Optimization for Language Model Alignment [75.83359213697854]
  Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences. We propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium.
  arXiv  Detail & Related papers  (2024-05-01T17:59:20Z)
• Edinburgh Clinical NLP at SemEval-2024 Task 2: Fine-tune your model unless you have access to GPT-4 [10.01547158445743]
  We evaluate various Large Language Models (LLMs) with multiple strategies, including Chain-of-Thought, In-Context Learning, and Efficient Fine-Tuning (PEFT) We found that the two PEFT adapters improves the F1 score (+0.0346) and consistency (+0.152) of the LLMs. Averaging the three metrics, GPT-4 ranks joint-first in the competition with 0.8328.
  arXiv  Detail & Related papers  (2024-03-30T22:27:21Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.