GRPO updates reduce to a damped oscillator whose mass, damping, and stiffness are fixed by optimizer hyperparameters plus one measured curvature scale, subsuming single-exponential saturation while adding inertial slow-start and group-size predictions.
Title resolution pending
18 Pith papers cite this work, alongside 2 external citations. Polarity classification is still indexing.
18
Pith papers citing it
2
external citations · external index
representative citing papers
ShredBench shows state-of-the-art MLLMs perform well on intact documents but suffer sharp drops in restoration accuracy as fragmentation increases to 8-16 pieces, indicating insufficient cross-modal semantic reasoning for VRDU.
A single LLM improves its own reasoning by self-distilling from privileged verified traces as teacher to its question-only student policy, outperforming off-policy distillation and RL on math benchmarks with better token efficiency.
An LLM agent self-evolves a set of query-rewriting rules that raise BM25 performance on the LeCaRD-v2 legal retrieval benchmark above human-designed and greedy baselines.
PROPEL amortizes solver evaluation with a trained activation probe to optimize task generators toward a target solve rate, raising the share of learnable tasks from ~10% to ~20% in coding and SWE experiments.
Using a 1PL IRT model on real cultural questions across 13 locales, the study identifies a local-language knowledge-access advantage masked by lower proficiency in raw accuracy.
SSOPD converts intra-group correct-wrong contrast into process supervision by distilling a teacher distribution from the shortest correct completion into prefixes of the longest wrong completion, improving GRPO on AIME and HMMT benchmarks.
Strong LLM optimizers act as local refiners with incremental improvements and semantic localization, while weaker ones show large drift and stagnation; solution novelty predicts success only when searches stay localized around high-performing regions.
Across four frontier reasoning models, 61–93% of correct chain-of-thought steps are redundant, and this over-thinking is provably optimal under any length-agnostic outcome reward.
PDDL planning problems are used to generate about one million precise reasoning steps for training Process Reward Models, and adding this data to existing datasets improves LLM performance on both mathematical and non-mathematical reasoning benchmarks.
Recovering an orthogonal basis from model activations yields a model-native skill characterization that improves reasoning Pass@1 by up to 41% via targeted data selection and supports inference steering, outperforming human-characterized alternatives.
Red-Bandit adapts online to LLM failure modes by dynamically selecting among RL-trained LoRA attack-style experts via a bandit policy, reporting SOTA ASR@10 on AdvBench with lower-perplexity prompts.
Across 15 LLMs, first-answer accuracy on arithmetic procedural execution falls from 63% at 5 steps to 20% at 95 steps, with under-execution increasing.
Treating language as a latent variable via polyGRPO RL improves Qwen2.5-7B-Instruct by 6.72% on English reasoning benchmarks and 6.89% on multilingual ones, with cross-task gains on commonsense reasoning from math-only training.
Task-reward reinforcement learning yields robust gains on math benchmarks for models like Llama-3.2-3B while distribution sharpening alone delivers only limited and unstable improvements.
Large reasoning models exhibit reasoning collapse, with accuracy dropping sharply beyond task-specific complexity thresholds in controlled versions of nine classical reasoning tasks using strict validity validators.
Ministral 3 releases 3B/8B/14B parameter-efficient language models with base, instruction, and reasoning variants derived via iterative pruning and distillation, including image understanding capabilities.
A literature synthesis that organizes post-training reasoning data research around data objects, usefulness factors, construction methods, and scaling behaviors to create an attribution framework.
citing papers explorer
-
Predictable GRPO: A Closed-Form Model of Training Dynamics
GRPO updates reduce to a damped oscillator whose mass, damping, and stiffness are fixed by optimizer hyperparameters plus one measured curvature scale, subsuming single-exponential saturation while adding inertial slow-start and group-size predictions.
-
ShredBench: Evaluating the Semantic Reasoning Capabilities of Multimodal LLMs in Document Reconstruction
ShredBench shows state-of-the-art MLLMs perform well on intact documents but suffer sharp drops in restoration accuracy as fragmentation increases to 8-16 pieces, indicating insufficient cross-modal semantic reasoning for VRDU.
-
Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models
A single LLM improves its own reasoning by self-distilling from privileged verified traces as teacher to its question-only student policy, outperforming off-policy distillation and RL on math benchmarks with better token efficiency.
-
When Rules Learn: A Self-Evolving Agent for Legal Case Retrieval
An LLM agent self-evolves a set of query-rewriting rules that raise BM25 performance on the LeCaRD-v2 legal retrieval benchmark above human-designed and greedy baselines.
-
Breaking the Solver Bottleneck: Training Task Generators at the Learnable Frontier
PROPEL amortizes solver evaluation with a trained activation probe to optimize task generators toward a target solve rate, raising the share of learnable tasks from ~10% to ~20% in coding and SWE experiments.
-
The Masked Advantage: Uncovering Local-Language Access to Cultural Knowledge in LLMs
Using a 1PL IRT model on real cultural questions across 13 locales, the study identifies a local-language knowledge-access advantage masked by lower proficiency in raw accuracy.
-
Self-Supervised On-Policy Distillation for Reasoning Language Models
SSOPD converts intra-group correct-wrong contrast into process supervision by distilling a teacher distribution from the shortest correct completion into prefixes of the longest wrong completion, improving GRPO on AIME and HMMT benchmarks.
-
What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search
Strong LLM optimizers act as local refiners with incremental improvements and semantic localization, while weaker ones show large drift and stagnation; solution novelty predicts success only when searches stay localized around high-performing regions.
-
How Much Thinking is Enough? Quantifying and Understanding Redundancy in LLM Reasoning
Across four frontier reasoning models, 61–93% of correct chain-of-thought steps are redundant, and this over-thinking is provably optimal under any length-agnostic outcome reward.
-
Process Reward Models Meet Planning: Generating Precise and Scalable Datasets for Step-Level Rewards
PDDL planning problems are used to generate about one million precise reasoning steps for training Process Reward Models, and adding this data to existing datasets improves LLM performance on both mathematical and non-mathematical reasoning benchmarks.
-
Characterizing Model-Native Skills
Recovering an orthogonal basis from model activations yields a model-native skill characterization that improves reasoning Pass@1 by up to 41% via targeted data selection and supports inference steering, outperforming human-characterized alternatives.
-
Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts
Red-Bandit adapts online to LLM failure modes by dynamically selecting among RL-trained LoRA attack-style experts via a bandit policy, reporting SOTA ASR@10 on AdvBench with lower-perplexity prompts.
-
When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models
Across 15 LLMs, first-answer accuracy on arithmetic procedural execution falls from 63% at 5 steps to 20% at 95 steps, with under-execution increasing.
-
Language as a Latent Variable for Reasoning Optimization
Treating language as a latent variable via polyGRPO RL improves Qwen2.5-7B-Instruct by 6.72% on English reasoning benchmarks and 6.89% on multilingual ones, with cross-task gains on commonsense reasoning from math-only training.
-
Beyond Distribution Sharpening: The Importance of Task Rewards
Task-reward reinforcement learning yields robust gains on math benchmarks for models like Llama-3.2-3B while distribution sharpening alone delivers only limited and unstable improvements.
-
Empirical Evidence of Complexity-Induced Limits in Large Language Models on Finite Discrete State-Space Problems with Explicit Validity Constraints
Large reasoning models exhibit reasoning collapse, with accuracy dropping sharply beyond task-specific complexity thresholds in controlled versions of nine classical reasoning tasks using strict validity validators.
-
Ministral 3
Ministral 3 releases 3B/8B/14B parameter-efficient language models with base, instruction, and reasoning variants derived via iterative pruning and distillation, including image understanding capabilities.
-
A Primer in Post-Training Reasoning Data: What We Know About How It Works
A literature synthesis that organizes post-training reasoning data research around data objects, usefulness factors, construction methods, and scaling behaviors to create an attribution framework.