Small well-performing LVLMs gain the largest benefits from test-time scaling (up to ~30% improvement), often matching or exceeding larger models, while visual tokens contribute mainly early in the reasoning chain.
hub Mixed citations
Diverse demonstrations improve in-context compositional generalization
Mixed citation behavior. Most common role is background (50%).
17
Pith papers citing it
Background
50% of classified citations
hub tools
citation-role summary
background 4
baseline 2
dataset 1
other 1
citation-polarity summary
representative citing papers
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.
TOKI types four common contradiction-resolution heuristics as bitemporal operators on a dual-row schema, supplies soundness theorems, and shows via a verdict matrix that it alone avoids three write-time anomalies while retaining a language-model judge.
A test-time zeroth-order optimization of prompt embeddings using a bounded self-supervised proxy from demonstration log-probabilities improves ICL accuracy and correlates with gains across tasks.
CoCoReviewBench curates 3,900 ICLR and NeurIPS papers into category-specific subsets with discussion-based annotations to evaluate AI reviewers on completeness and correctness rather than human review overlap.
PuzzleMark provides a robust and imperceptible watermarking method for code datasets using adaptive variable name concatenation and statistical verification, achieving perfect detection rates with minimal performance impact.
DiLaServe improves SLO attainment for diffusion language models by up to 56.6 percentage points and reduces latency by up to 46% with less than 1% accuracy drop via deadline-aware scheduling and dynamic reconfiguration.
SLM adds a dedicated spatial modality and training dataset to LLMs, enabling geometric spatial reasoning and outperforming prompt-based symbolic methods on the new SpatialEval benchmark.
A distribution-correction framework for offline LLM reasoning distillation improves accuracy on math benchmarks by adaptively aligning teacher supervision with the student's inference-time distribution.
An endpoint point-relation classifier followed by decoding to interval relations achieves 70.1% temporal awareness on TempEval-3, setting a new state-of-the-art for the full set of fine-grained temporal relations.
Re-evaluating controlled text generation systems under standardized conditions reveals that many published performance claims do not hold, highlighting the need for consistent evaluation practices.
LLMs produce overly positive idealized depictions of disability in simulated social media posts that do not match real posts by people with disabilities and show topic bias favoring nondisabled people.
A literature review concludes that pursuing consensus in data annotation creates biased AI by dismissing subjective disagreements and enforcing geographic hegemony, and proposes mapping diversity instead.
Mujica-MyGo decomposes multi-turn RAG interactions via multi-agent workflows and applies minimalist policy gradient optimization to improve performance on QA benchmarks while avoiding long-context problems.
Mimir is a 1.6B multilingual concept model pretrained on 38.9 billion sentences across 46 languages and instruction-tuned on 66.8 million sentences across 35 languages, then compared to a token-based LM of similar size.
Secondary analysis of 30 Saudi Google users' interviews identifies balanced perceptions of activity logs spanning benefits, risks, misconceptions, and severe negative outcomes.
A survey reviewing benchmark data contamination in LLMs, its impact on evaluation, and alternative assessment approaches.
citing papers explorer
-
On Test-Time Scaling for Vision-Language Models
Small well-performing LVLMs gain the largest benefits from test-time scaling (up to ~30% improvement), often matching or exceeding larger models, while visual tokens contribute mainly early in the reasoning chain.
-
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.
-
TOKI: A Bitemporal Operator Algebra for Contradiction Resolution in LLM-Agent Persistent Memory
TOKI types four common contradiction-resolution heuristics as bitemporal operators on a dual-row schema, supplies soundness theorems, and shows via a verdict matrix that it alone avoids three write-time anomalies while retaining a language-model judge.
-
Self-Improving In-Context Learning
A test-time zeroth-order optimization of prompt embeddings using a bounded self-supervised proxy from demonstration log-probabilities improves ICL accuracy and correlates with gains across tasks.
-
CoCoReviewBench: A Completeness- and Correctness-Oriented Benchmark for AI Reviewers
CoCoReviewBench curates 3,900 ICLR and NeurIPS papers into category-specific subsets with discussion-based annotations to evaluate AI reviewers on completeness and correctness rather than human review overlap.
-
PuzzleMark: Implicit Jigsaw Learning for Robust Code Dataset Watermarking in Neural Code Completion Models
PuzzleMark provides a robust and imperceptible watermarking method for code datasets using adaptive variable name concatenation and statistical verification, achieving perfect detection rates with minimal performance impact.
-
DiLaServe: High SLO Attainment Serving for Diffusion Language Models
DiLaServe improves SLO attainment for diffusion language models by up to 56.6 percentage points and reduces latency by up to 46% with less than 1% accuracy drop via deadline-aware scheduling and dynamic reconfiguration.
-
From Symbolic to Geometric: Enabling Spatial Reasoning in Large Language Models
SLM adds a dedicated spatial modality and training dataset to LLMs, enabling geometric spatial reasoning and outperforming prompt-based symbolic methods on the new SpatialEval benchmark.
-
Distribution Corrected Offline Data Distillation for Large Language Models
A distribution-correction framework for offline LLM reasoning distillation improves accuracy on math benchmarks by adaptively aligning teacher supervision with the student's inference-time distribution.
-
Looking for the Bottleneck in Fine-grained Temporal Relation Classification
An endpoint point-relation classifier followed by decoding to interval relations achieves 70.1% temporal awareness on TempEval-3, setting a new state-of-the-art for the full set of fine-grained temporal relations.
-
A Comparative Study of Controlled Text Generation Systems Using Level-Playing-Field Evaluation Principles
Re-evaluating controlled text generation systems under standardized conditions reveals that many published performance claims do not hold, highlighting the need for consistent evaluation practices.
-
Shiny Stories, Hidden Struggles: Investigating the Representation of Disability Through the Lens of LLMs
LLMs produce overly positive idealized depictions of disability in simulated social media posts that do not match real posts by people with disabilities and show topic bias favoring nondisabled people.
-
The Consensus Trap: Dissecting Subjectivity and the "Ground Truth" Illusion in Data Annotation
A literature review concludes that pursuing consensus in data annotation creates biased AI by dismissing subjective disagreements and enforcing geographic hegemony, and proposes mapping diversity instead.
-
Advancing Multi-Agent RAG Systems with Minimalist Reinforcement Learning
Mujica-MyGo decomposes multi-turn RAG interactions via multi-agent workflows and applies minimalist policy gradient optimization to improve performance on QA benchmarks while avoiding long-context problems.
-
Mimir: Large-scale Multilingual Concept Modeling
Mimir is a 1.6B multilingual concept model pretrained on 38.9 billion sentences across 46 languages and instruction-tuned on 66.8 million sentences across 35 languages, then compared to a token-based LM of similar size.
-
Users' Activity Logs: the Good, the Bad, the Misconception, and the Disastrous
Secondary analysis of 30 Saudi Google users' interviews identifies balanced perceptions of activity logs spanning benefits, risks, misconceptions, and severe negative outcomes.
-
Benchmark Data Contamination of Large Language Models: A Survey
A survey reviewing benchmark data contamination in LLMs, its impact on evaluation, and alternative assessment approaches.