Teaching large language models to reason with reinforcement learning

representative citing papers

citing papers explorer

Showing 17 of 17 citing papers.

• Logic-Regularized Verifier Elicits Reasoning from LLMs cs.CL · 2026-05-07 · unverdicted · none · ref 72

  LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.

• CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignment cs.SE · 2025-10-21 · conditional · none · ref 10

  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.

• Training Large Language Models to Reason in a Continuous Latent Space cs.CL · 2024-12-09 · unverdicted · none · ref 16

  Coconut lets LLMs perform reasoning directly in continuous latent space by recycling hidden states as inputs, outperforming standard chain-of-thought on search-intensive logical tasks with better accuracy-efficiency trade-offs.

• When Reasoning Traces Become Performative: Step-Level Evidence that Chain-of-Thought Is an Imperfect Oversight Channel cs.AI · 2026-05-12 · unverdicted · none · ref 22

  CoT traces align with internal answer commitment in only 61.9% of steps on average, dominated by confabulated continuations after commitment has stabilized.

• Epistemic Uncertainty for Test-Time Discovery cs.LG · 2026-05-11 · unverdicted · none · ref 11

  UG-TTT adds epistemic uncertainty measured by adapter disagreement as an exploration bonus in RL for LLMs, raising maximum reward and diversity on scientific discovery benchmarks.

• SeLaR: Selective Latent Reasoning in Large Language Models cs.CL · 2026-04-09 · unverdicted · none · ref 15

  SeLaR selectively applies latent soft reasoning in LLMs via entropy gating and contrastive regularization, outperforming standard CoT on five benchmarks without training.

• Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning cs.CL · 2025-12-08 · unverdicted · none · ref 19

  NPR trains LLMs to reason in parallel via self-distilled RL, delivering up to 24.5% performance gains and 4.6x speedups with 100% genuine parallel execution on reasoning benchmarks.

• Beyond Memorization: Extending Reasoning Depth with Recurrence, Memory and Test-Time Compute Scaling cs.LG · 2025-08-22 · unverdicted · none · ref 31

  In a cellular automata rule-inference task designed to block memorization, neural models achieve high next-step accuracy but accuracy falls sharply with longer reasoning chains; depth, recurrence, memory, and test-time compute extend the reachable depth but do not remove the bound.

• RL-PLUS: Countering Capability Boundary Collapse of LLMs in Reinforcement Learning with Hybrid-policy Optimization cs.AI · 2025-07-31 · unverdicted · none · ref 8

  RL-PLUS is a hybrid RL approach for LLMs that combines internal exploitation with external data via importance sampling and exploration advantages to prevent capability boundary collapse and achieve gains on math and OOD reasoning benchmarks.

• LoVeC: Reinforcement Learning for Better Verbalized Confidence in Long-Form Generations cs.CL · 2025-05-29 · unverdicted · none · ref 49

  LoVeC uses RL to train LLMs to output verbalized numerical confidence scores for statements in long-form text, achieving better calibration than self-consistency baselines on QA datasets while being 20x faster.

• Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning cs.LG · 2024-10-10 · unverdicted · none · ref 6

  Process advantage verifiers trained to predict step-level progress under a distinct prover policy improve LLM reasoning accuracy by over 8% and sample efficiency by 5-6x over outcome reward models.

• Training Language Models to Self-Correct via Reinforcement Learning cs.LG · 2024-09-19 · unverdicted · none · ref 7

  SCoRe uses multi-turn online RL with regularization on self-generated traces to improve LLM self-correction, achieving 15.6% and 9.1% gains on MATH and HumanEval for Gemini models.

• NoisyCoconut: Counterfactual Consensus via Latent Space Reasoning cs.LG · 2026-05-06 · unverdicted · none · ref 34

  Injecting noise into LLM latent trajectories creates diverse reasoning paths whose agreement acts as a confidence signal for selective abstention, cutting error rates from 40-70% to under 15% on math tasks.

• Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding cs.LG · 2026-04-23 · unverdicted · none · ref 30

  A co-evolving proposer-critic RL framework improves GUI grounding accuracy by letting the model critique its own proposals rendered on screenshots.

• Internalizing Outcome Supervision into Process Supervision: A New Paradigm for Reinforcement Learning for Reasoning cs.LG · 2026-04-19 · unverdicted · none · ref 4

  A method internalizes outcome supervision into process supervision by extracting step-level learning signals from failed reasoning trajectories during reinforcement learning.

• Learning to Reason at the Frontier of Learnability cs.LG · 2025-02-17 · unverdicted · none · ref 26

  A curriculum sampling questions with high variance in success rate improves reinforcement learning performance for LLM reasoning tasks.

• LiFT: How to Enable In-Context Learning for Longitudinal Modelling cs.CL · 2026-03-25 · unreviewed · ref 2