CoT transformers simulate any Word RAM algorithm with poly-logarithmic overhead in three architectures, improving on quadratic TM overhead.
Mixed citations
Title resolution pending
Mixed citation behavior. Most common role is background (68%).
224
Pith papers citing it
539
external citations · Crossref
Background
68% of classified citations
citation-role summary
background 26
method 5
dataset 3
baseline 2
other 1
citation-polarity summary
co-cited works
representative citing papers
Across 30 LLMs and 205 TLA+ tasks, syntactic correctness reaches at most 26.6% and semantic correctness 8.6%, with all successes limited to progressive prompting and no advantage from larger models.
Introduces BonaFide benchmark of 3,066 ground-truth labeled CoTs showing most faithfulness metrics perform near chance with biases and poor scaling to longer chains.
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
Looped transformers with recall and outer normalization produce reachable, input-dependent fixed points with stable gradients, enabling generalization, while those without recall cannot; a new internal recall variant performs competitively or better.
The SDE benchmark shows LLMs lag on scientific discovery tasks relative to general science tests, with diminishing scaling returns and shared weaknesses across models.
SwissGov-RSD is the first naturalistic cross-lingual document-level benchmark with human token-level semantic difference annotations, on which both LLMs and encoders show a large performance gap relative to simpler settings.
SimBench unifies 20 datasets into the first large-scale benchmark, finding top LLMs reach only modest human simulation fidelity of 40.8/100 with log-linear scaling by size and an alignment tradeoff on diverse questions.
Spurious rewards in RLVR can produce large gains in mathematical reasoning for certain language models via GRPO's clipping bias amplifying pretraining behaviors like code reasoning.
Controlled student-teacher experiments across four benchmarks show interactive gains are driven more by the student's ability to use feedback than by teacher quality, with self-feedback adding little beyond unguided retries.
DLR creates discrete latent tokens from rendered CoT images via clustering, enabling up to 20x compression and interpretable trajectories that outperform continuous latent baselines on reasoning tasks.
ATHENA-R1 is an RL-trained agent using 212 biomedical tools that achieves 94.7% accuracy on drug reasoning and 82.9% on treatment reasoning tasks, outperforming GPT-5 by 17.8 and 10.7 points respectively.
MRI2Rep generates LI-RADS structured reports from 3D liver MRI via autoregressive modeling on 3929 real-world pairs, reporting 76% case-level sensitivity and 70-75% clinical acceptability in reader study.
SGPO extracts strategies from strong-model responses, builds autonomous and guided trajectories, and applies token-level forward-KL distillation with adaptive weighting to outperform SFT and RL baselines by 2.2 points on math benchmarks.
SC-GRPO improves RL with verifiable rewards by multiplying GRPO gradients with self-induced per-token KL divergence, outperforming GRPO by 8.1% and DAPO by 5.9% on math, code, and agent benchmarks.
PowerOPD applies the Box-Cox power transformation to create natively bounded, sign-consistent rewards for on-policy distillation, delivering up to +6.37 Avg@8 gains over vanilla OPD on math reasoning benchmarks while cutting compute costs.
MetaSyn benchmark shows LLM agents recover at most 52.7% of relevant studies in meta-analysis pipelines due to failures in PI/ECO-based screening despite strong retrieval.
RGSD distills rubric-conditioned teacher distributions into base policies token-by-token, matching GRPO rubric satisfaction on Qwen models with one rollout and zero verifier calls.
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
A stop-gradient consistency regularizer mitigates context-induced degradation in on-policy distillation, improving robustness across 12 configurations.
Optical reasoning encodes rationales in images rather than text, matching or exceeding text-based performance on math, science, and multimodal benchmarks while cutting tokens by 28.57% on language tasks and 16% on multimodal tasks.
RL with chrF reward trains LLMs to better utilize in-context linguistic knowledge for zero-shot translation of unseen languages, outperforming ICL and SFT.
LatentSkill uses a hypernetwork to generate LoRA adapters from textual skills, enabling weight-space storage that cuts prefill tokens and boosts agent success rates on ALFWorld and Search-QA.
ACE-SQL jointly optimizes schema linking and SQL generation via RL with empirical credit assignment from execution-correct rollouts, achieving 65.3% greedy execution accuracy on BIRD Dev using 0.93k output tokens.
citing papers explorer
-
Efficiently Representing Algorithms With Chain-of-Thought Transformers
CoT transformers simulate any Word RAM algorithm with poly-logarithmic overhead in three architectures, improving on quadratic TM overhead.
-
Can LLMs Write Correct TLA+ Specifications? Evaluating Natural-Language-to-TLA+ Generation
Across 30 LLMs and 205 TLA+ tasks, syntactic correctness reaches at most 26.6% and semantic correctness 8.6%, with all successes limited to progressive prompting and no advantage from larger models.
-
Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth
Introduces BonaFide benchmark of 3,066 ground-truth labeled CoTs showing most faithfulness metrics perform near chance with biases and poor scaling to longer chains.
-
Efficient Training on Multiple Consumer GPUs with RoundPipe
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
-
Stability and Generalization in Looped Transformers
Looped transformers with recall and outer normalization produce reachable, input-dependent fixed points with stable gradients, enabling generalization, while those without recall cannot; a new internal recall variant performs competitively or better.
-
Evaluating Large Language Models in Scientific Discovery
The SDE benchmark shows LLMs lag on scientific discovery tasks relative to general science tests, with diminishing scaling returns and shared weaknesses across models.
-
SwissGov-RSD: A Human-annotated, Cross-lingual Benchmark for Token-level Recognition of Semantic Differences Between Related Documents
SwissGov-RSD is the first naturalistic cross-lingual document-level benchmark with human token-level semantic difference annotations, on which both LLMs and encoders show a large performance gap relative to simpler settings.
-
SimBench: Benchmarking the Ability of Large Language Models to Simulate Human Behaviors
SimBench unifies 20 datasets into the first large-scale benchmark, finding top LLMs reach only modest human simulation fidelity of 40.8/100 with log-linear scaling by size and an alignment tradeoff on diverse questions.
-
Spurious Rewards: Rethinking Training Signals in RLVR
Spurious rewards in RLVR can produce large gains in mathematical reasoning for certain language models via GRPO's clipping bias amplifying pretraining behaviors like code reasoning.
-
What Drives Interactive Improvement from Feedback?
Controlled student-teacher experiments across four benchmarks show interactive gains are driven more by the student's ability to use feedback than by teacher quality, with self-feedback adding little beyond unguided retries.
-
Why Struggle with Continuous Latents? Interpretable Discrete Latent Reasoning via Rendered Compression
DLR creates discrete latent tokens from rendered CoT images via clustering, enabling up to 20x compression and interpretable trajectories that outperform continuous latent baselines on reasoning tasks.
-
An AI agent for treatment reasoning over a biomedical tool universe
ATHENA-R1 is an RL-trained agent using 212 biomedical tools that achieves 94.7% accuracy on drug reasoning and 82.9% on treatment reasoning tasks, outperforming GPT-5 by 17.8 and 10.7 points respectively.
-
MRI2Rep: Autoregressive Structured Report Generation for 3D Liver MRI
MRI2Rep generates LI-RADS structured reports from 3D liver MRI via autoregressive modeling on 3929 real-world pairs, reporting 76% case-level sensitivity and 70-75% clinical acceptability in reader study.
-
Beyond Trajectory Imitation: Strategy-Guided Policy Optimization for LLM Reasoning
SGPO extracts strategies from strong-model responses, builds autonomous and guided trajectories, and applies token-level forward-KL distillation with adaptive weighting to outperform SFT and RL baselines by 2.2 points on math benchmarks.
-
Learning from Own Solutions: Self-Conditioned Credit Assignment for Reinforcement Learning with Verifiable Rewards
SC-GRPO improves RL with verifiable rewards by multiplying GRPO gradients with self-induced per-token KL divergence, outperforming GRPO by 8.1% and DAPO by 5.9% on math, code, and agent benchmarks.
-
PowerOPD: Stabilizing On-Policy Distillation with Bounded Power Transformation
PowerOPD applies the Box-Cox power transformation to create natively bounded, sign-consistent rewards for on-policy distillation, delivering up to +6.37 Avg@8 gains over vanilla OPD on math reasoning benchmarks while cutting compute costs.
-
Benchmarking LLM Agents on Meta-Analysis Articles from Nature Portfolio
MetaSyn benchmark shows LLM agents recover at most 52.7% of relevant studies in meta-analysis pipelines due to failures in PI/ECO-based screening despite strong retrieval.
-
Rubric-Guided Self-Distillation: Post-Training Without Rubric Verifiers
RGSD distills rubric-conditioned teacher distributions into base policies token-by-token, matching GRPO rubric satisfaction on Qwen models with one rollout and zero verifier calls.
-
Fine-tuning Multi-modal LLMs with ART: Art-based Reinforcement Training
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
-
When Context Returns: Toward Robust Internalization in On-Policy Distillation
A stop-gradient consistency regularizer mitigates context-induced degradation in on-policy distillation, improving robustness across 12 configurations.
-
Optical Reasoning: Rethinking Images as an Expressive Reasoning Medium Beyond Text
Optical reasoning encodes rationales in images rather than text, matching or exceeding text-based performance on math, science, and multimodal benchmarks while cutting tokens by 28.57% on language tasks and 16% on multimodal tasks.
-
Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation
RL with chrF reward trains LLMs to better utilize in-context linguistic knowledge for zero-shot translation of unseen languages, outperforming ICL and SFT.
-
LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents
LatentSkill uses a hypernetwork to generate LoRA adapters from textual skills, enabling weight-space storage that cuts prefill tokens and boosts agent success rates on ALFWorld and Search-QA.
-
ACE-SQL: Adaptive Co-Optimization via Empirical Credit Assignment for Text-to-SQL
ACE-SQL jointly optimizes schema linking and SQL generation via RL with empirical credit assignment from execution-correct rollouts, achieving 65.3% greedy execution accuracy on BIRD Dev using 0.93k output tokens.
-
ATLAS: Agentic Test-time Learning-to-Allocate Scaling
ATLAS introduces an LLM-orchestrated agentic framework for dynamic test-time scaling via extensible 'explore' actions, achieving higher accuracy with fewer API calls than fixed-workflow baselines on four benchmarks.
-
Reasoning with Sampling: Cutting at Decision Points
Entropy-Cut Metropolis-Hastings targets high-entropy decision points for resampling, yielding mixing time that scales with the number of decisions and consistent gains over baselines on MATH500, HumanEval, GPQA Diamond, and AIME26.
-
Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL
Extrapolative weight averaging of RL checkpoints trained under nested unit-test coverage extends a correctness-efficiency frontier and boosts ensemble pass rates in code generation across model scales and inference modes.
-
Do Agents Think Deeper? A Mechanistic Investigation of Layer-Wise Dynamics in Sequential Planning
LLM agents progressively recruit deeper layers with stronger long-range dependencies and correction-dominant residual updates during sequential planning, showing a construction-refinement gap unlike static tasks.
-
Conceptual Steganography
Conceptual steganography encodes covert information in high-level reasoning patterns within LM chains-of-thought, remaining robust to paraphrase defenses while preserving reasoning utility.
-
CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents
CUA-Gym generates 32,112 verified RLVR tuples across 110 mock environments, enabling trained models to reach 62.1% and 72.6% on OSWorld-Verified while transferring to WebArena.
-
Reasoning Portability: Guiding Continual Learning for MLLMs in the RLVR Era
Formalizes Reasoning Portability (RP) and proposes RDB-CL to modulate per-sample KL regularization in RLVR for MLLM continual learning, achieving +12.0% Last accuracy over vanilla RLVR baseline by preserving reusable reasoning on high-RP samples.
-
Learning, Fast and Slow: Towards LLMs That Adapt Continually
Fast-Slow Training uses context optimization as fast weights alongside parameter updates as slow weights to achieve up to 3x better sample efficiency, higher performance, and less catastrophic forgetting than standard RL in continual LLM learning.
-
Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning
RAPO uses an information-theoretic lower bound on visual gain to select high-entropy reflection anchors and optimizes a chain-masked KL surrogate, delivering gains over baselines on reasoning benchmarks across LVLM backbones.
-
UniShield: Unified Face Attack Detection via KG-Informed Multimodal Reasoning
UniShield introduces a knowledge-graph-informed multimodal framework that improves unified detection of physical and digital face attacks through instruction tuning and consistency-optimized reasoning.
-
Tracing Uncertainty in Language Model "Reasoning"
Uncertainty trace profiles from LM reasoning traces predict correct final answers with AUROC up to 0.807 and enable early error detection using only initial tokens.
-
Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States
POISE trains a lightweight probe on the actor's internal states to predict expected rewards for RLVR, matching DAPO performance on math benchmarks with lower compute by avoiding extra rollouts or critic models.
-
LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification
LaTER reduces LLM token usage 16-33% on reasoning benchmarks by exploring in latent space then switching to explicit CoT verification, with gains like 70% to 73.3% on AIME 2025 in the training-free version.
-
TableVista: Benchmarking Multimodal Table Reasoning under Visual and Structural Complexity
TableVista benchmark finds foundation models maintain performance across visual styles but degrade sharply on complex table structures and vision-only settings.
-
Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration
Prepending stochastic sequences from Lorem Ipsum vocabulary to prompts during GRPO resampling broadens reasoning exploration and outperforms standard resampling on hard tasks for 1.7B-7B models.
-
RAG over Thinking Traces Can Improve Reasoning Tasks
Retrieving structured thinking traces as a corpus improves reasoning performance on AIME, LiveCodeBench, and GPQA over standard RAG or no retrieval.
-
DiagramNet: An End-to-End Recognition Framework and Dataset for Non-Standard System-Level Diagrams
DiagramNet supplies a new multimodal dataset and progressive training pipeline with decoupled multi-agent workflow, allowing a 3B model to outperform GPT-5, Claude-Sonnet-4, and Gemini-2.5-Pro by over 2x on system-level diagram tasks while generalizing to other benchmarks.
-
BoostLoRA: Growing Effective Rank by Boosting Adapters
BoostLoRA grows effective adapter rank linearly via iterative boosting on hard examples with orthogonal low-rank updates, outperforming both single-shot ultra-low-rank adapters and full fine-tuning on math and code tasks with zero added inference overhead.
-
When to Retrieve During Reasoning: Adaptive Retrieval for Large Reasoning Models
ReaLM-Retrieve uses step-level uncertainty to trigger retrievals during reasoning, achieving 10.1% better F1 scores and 47% fewer calls on multi-hop QA benchmarks.
-
TRIP-Evaluate: An Open Multimodal Benchmark for Evaluating Large Models in Transportation
TRIP-Evaluate is a new open multimodal benchmark with 837 text, image, and point-cloud items organized by a role-task-knowledge taxonomy to evaluate large models on transportation workflows.
-
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
Abstract-CoT lets models reason with short discrete latent token sequences from a reserved vocabulary, using warm-up training and RL to match verbal CoT performance with up to 11.6x fewer tokens.
-
EVPO: Explained Variance Policy Optimization for Adaptive Critic Utilization in LLM Post-Training
EVPO adaptively switches between critic-based and batch-mean advantage estimation using batch-level explained variance to provably achieve no greater variance than the better of PPO or GRPO at every step.
-
Wiring the 'Why': A Unified Taxonomy and Survey of Abductive Reasoning in LLMs
The paper delivers the first survey of abductive reasoning in LLMs, a unified two-stage taxonomy, a compact benchmark, and an analysis of gaps relative to deductive and inductive reasoning.
-
Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning
This survey introduces the Generate-Filter-Control-Replay (GFCR) taxonomy to structure rollout pipelines for RL-based post-training of reasoning LLMs.
-
MMORF: A Multi-agent Framework for Designing Multi-objective Retrosynthesis Planning Systems
MMORF provides a modular multi-agent framework for multi-objective retrosynthesis planning, with MASIL and RFAS systems showing strong safety, cost, and success metrics on a new 218-task benchmark.
-
The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?
The Stepwise Informativeness Assumption explains the correlation between LLM entropy dynamics and reasoning correctness by positing that correct traces accumulate answer-relevant information stepwise during generation.