A survey of 1,250 papers organizes AI self-improvement along two axes—what is improved and loop closure—finding that demonstrated self-improvement strength tracks a verification hierarchy from formal verifiers down to intrinsic self-assessment.
Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
abstract
Large language models (LLMs) have shown impressive capabilities, but still struggle with complex reasoning tasks requiring multiple steps. While prompt-based methods like Chain-of-Thought (CoT) can improve LLM reasoning at inference time, optimizing reasoning capabilities during training remains challenging. We introduce LaTent Reasoning Optimization (LaTRO), a principled framework that formulates reasoning as sampling from a latent distribution and optimizes it via variational approaches. LaTRO enables LLMs to concurrently improve both their reasoning process and ability to evaluate reasoning quality, without requiring external feedback or reward models. We validate LaTRO through experiments on GSM8K and ARC-Challenge datasets using multiple model architectures. On GSM8K, LaTRO improves zero-shot accuracy by an average of 12.5% over base models and 9.6% over supervised fine-tuning across Phi-3.5-mini, Mistral-7B, and Llama-3.1-8B. Our findings suggest that pre-trained LLMs possess latent reasoning capabilities that can be unlocked and enhanced through our proposed optimization approach in a self-improvement manner. The code of LaTRO is available at \url{https://github.com/SalesforceAIResearch/LaTRO}.
citation-role summary
background 2
other 1
citation-polarity summary
representative citing papers
Low-rank graphs induce latents that form causal abstractions, with identifiability results and a practical objective enabling unsupervised learning of high-level SCMs from low-level measurements.
Latent reasoning through eight bottleneck tokens gives multimodal retrieval performance close to explicit chain-of-thought models at a fraction of the inference cost.
LASAR uses two-stage supervised training plus reinforcement learning to ground semantic IDs, align latent reasoning trajectories to CoT hidden states via KL divergence, and adaptively choose reasoning depth, halving average steps while improving quality on three datasets.
R&B-EnCoRe uses self-supervised importance-weighted variational inference to distill action-predictive reasoning datasets that improve VLA performance on manipulation, navigation, and driving tasks without external verifiers.
Higher sequence probability predicts correctness across different answers in a dataset but does not reliably improve accuracy when decoding methods or hyperparameters are changed, nor does it indicate correctness for repeated responses to one prompt.
VI-CuRL stabilizes verifier-independent RL for LLM reasoning via confidence-guided curriculum that reduces action and problem variance, with a claimed proof of asymptotic unbiasedness and empirical gains over baselines.
Direct Reasoning Optimization applies token-level Reasoning Reflection Reward (R3) focused on high-variance tokens and rubric-gating constraints to improve sample-efficient RL training of LLMs on unverifiable tasks.
The paper unifies perspectives on Long CoT in reasoning LLMs by introducing a taxonomy, detailing characteristics of deep reasoning and reflection, and discussing emergence phenomena and future directions.
citing papers explorer
-
Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops
A survey of 1,250 papers organizes AI self-improvement along two axes—what is improved and loop closure—finding that demonstrated self-improvement strength tracks a verification hierarchy from formal verifiers down to intrinsic self-assessment.
-
Unsupervised Causal Abstractions Discovery
Low-rank graphs induce latents that form causal abstractions, with identifiability results and a practical objective enabling unsupervised learning of high-level SCMs from low-level measurements.
-
LaME: Learning to Think in Latent Space for Multimodal Embedding via Information Bottleneck
Latent reasoning through eight bottleneck tokens gives multimodal retrieval performance close to explicit chain-of-thought models at a fraction of the inference cost.
-
LASAR: Latent Adaptive Semantic Aligned Reasoning for Generative Recommendation
LASAR uses two-stage supervised training plus reinforcement learning to ground semantic IDs, align latent reasoning trajectories to CoT hidden states via KL divergence, and adaptively choose reasoning depth, halving average steps while improving quality on three datasets.
-
Self-Supervised Bootstrapping of Action-Predictive Embodied Reasoning
R&B-EnCoRe uses self-supervised importance-weighted variational inference to distill action-predictive reasoning datasets that improve VLA performance on manipulation, navigation, and driving tasks without external verifiers.
-
When are likely answers right? On Sequence Probability and Correctness in LLMs
Higher sequence probability predicts correctness across different answers in a dataset but does not reliably improve accuracy when decoding methods or hyperparameters are changed, nor does it indicate correctness for repeated responses to one prompt.
-
VI-CuRL: Stabilizing Verifier-Independent RL Reasoning via Confidence-Guided Variance Reduction
VI-CuRL stabilizes verifier-independent RL for LLM reasoning via confidence-guided curriculum that reduces action and problem variance, with a claimed proof of asymptotic unbiasedness and empirical gains over baselines.
-
Direct Reasoning Optimization: Token-Level Reasoning Reflectivity Meets Rubric Gates for Unverifiable Tasks
Direct Reasoning Optimization applies token-level Reasoning Reflection Reward (R3) focused on high-variance tokens and rubric-gating constraints to improve sample-efficient RL training of LLMs on unverifiable tasks.
-
Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models
The paper unifies perspectives on Long CoT in reasoning LLMs by introducing a taxonomy, detailing characteristics of deep reasoning and reflection, and discussing emergence phenomena and future directions.