HLS-Seek replaces full-synthesis RL with a comparative proxy reward model plus uncertainty-triggered real checks, yielding higher correctness and better QoR than larger models at 8.5x lower training cost.
Canonical reference
Title resolution pending
Canonical reference. 100% of citing Pith papers cite this work as background.
38
Pith papers citing it
Background
100% of classified citations
citation-role summary
background 6
citation-polarity summary
roles
background 6polarities
background 6representative citing papers
MLLMs display a large perception-reasoning gap on perspective-conditioned spatial reasoning tasks from omnidirectional images, with sharp accuracy drops on advanced tasks like egocentric rotation, though partial gains are possible via RL reward shaping.
Misrouter enables input-only attacks on MoE LLMs by optimizing queries on open-source surrogates to route toward weakly aligned experts and transferring them to public APIs.
RouteHijack is a routing-aware jailbreak that identifies safety-critical experts via activation contrast and optimizes suffixes to suppress them, reaching 69.3% average attack success rate on seven MoE LLMs with strong transfer to variants and VLMs.
RLHF should decompose annotations into dimensions each matched to one of three models—extension, evidence, or authority—instead of applying a single unified pipeline.
PUDA enables effective promotion of unpopular target items in black-box LLM sequential recommenders by using evolutionary LLM refinement to infer hidden prompts, training a surrogate model, and combining adversarial text revision with surrogate-generated poisoning sequences.
SAGE is a new multi-agent benchmark that formalizes service SOPs as dynamic dialogue graphs to measure LLM agents on logical compliance and path coverage, uncovering an execution gap and empathy resilience across 27 models in 6 scenarios.
GeoSkill lets vision-language models improve geolocation accuracy and reasoning by maintaining an evolving Skill-Graph that grows through autonomous analysis of successful and failed rollouts on web-scale image data.
ToolPRM provides fine-grained intra-call process supervision via a new dataset and reward model, outperforming outcome and coarse-grained alternatives on function-calling benchmarks.
Semantic Compliance Hijacking lets attackers hijack LLM agents by disguising malicious instructions as compliance rules in skills, reaching up to 77.67% success on confidentiality breaches and 67.33% on RCE while evading all tested scanners.
AgentGR uses semantic-aware LLM agents to simulate group decision dynamics and improve group recommendation accuracy over traditional aggregation methods.
Introduces VURB benchmark and VUP-35K dataset to train discriminative and generative video reward models that achieve SOTA performance on VURB and VideoRewardBench.
A methodological framework and browser system BITE for collecting evolving user preferences on LLM outputs through context-triggered reflections and privacy-preserving data over time.
LocalAlign generates near-target adversarial examples via prompting and applies margin-aware alignment training to enforce tighter boundaries against prompt injection attacks.
CoRM-RAG uses a cognitive perturbation protocol to simulate biases and trains an Evidence Critic to retrieve documents that support correct decisions even under adversarial query changes.
Ethics testing is introduced as a systematic approach to generate tests that identify software harms induced by unethical behavior in generative AI outputs.
CAT improves line coverage by 18% and branch coverage by 22% over prior LLM test generation methods by adding call-chain and dependency context from static analysis to prompts.
GeoMind applies an agentic workflow with tool-augmented modules and process supervision to outperform static models on lithology classification from well logs while producing traceable decisions.
Infection-Reasoner, a 4B VLM, reaches 86.8% accuracy on wound infection classification while producing rationales rated mostly correct by experts, via GPT-5.1 distillation followed by reinforcement learning.
VB-Score shows three major LLMs have severe failures in medical entity recognition and factual consistency, with 13.8% lower performance on chronic conditions affecting older and minority groups, indicating condition-based algorithmic discrimination.
RoTRAG retrieves Rules of Thumb to ground LLM reasoning for harm detection and severity classification in multi-turn dialogues, reporting roughly 40% relative F1 gains and 8.4% lower distributional error on two safety benchmarks while cutting redundant retrieval.
Dual Reasoning with explicit safety audits improves the new SUDS metric by 1.32x to 3.42x over baselines on code generation benchmarks containing injected harmful keywords.
GRASP improves multimodal sarcasm target identification by anchoring visual regions in grounded chain-of-thought reasoning and using dual-stage optimization on a new balanced dataset.
VC-Soup uses a cosine-similarity consistency metric to filter data, trains value-consistent policies, and applies linear merging with Pareto filtering to improve multi-value LLM alignment trade-offs.
citing papers explorer
-
HLS-Seek: QoR-Aware Code Generation for High-Level Synthesis via Proxy Comparative Reward Reinforcement Learning
HLS-Seek replaces full-synthesis RL with a comparative proxy reward model plus uncertainty-triggered real checks, yielding higher correctness and better QoR than larger models at 8.5x lower training cost.
-
Beyond Localization: A Comprehensive Diagnosis of Perspective-Conditioned Spatial Reasoning in MLLMs from Omnidirectional Images
MLLMs display a large perception-reasoning gap on perspective-conditioned spatial reasoning tasks from omnidirectional images, with sharp accuracy drops on advanced tasks like egocentric rotation, though partial gains are possible via RL reward shaping.
-
Misrouter: Exploiting Routing Mechanisms for Input-Only Attacks on Mixture-of-Experts LLMs
Misrouter enables input-only attacks on MoE LLMs by optimizing queries on open-source surrogates to route toward weakly aligned experts and transferring them to public APIs.
-
RouteHijack: Routing-Aware Attack on Mixture-of-Experts LLMs
RouteHijack is a routing-aware jailbreak that identifies safety-critical experts via activation contrast and optimizes suffixes to suppress them, reaching 69.3% average attack success rate on seven MoE LLMs with strong transfer to variants and VLMs.
-
Three Models of RLHF Annotation: Extension, Evidence, and Authority
RLHF should decompose annotations into dimensions each matched to one of three models—extension, evidence, or authority—instead of applying a single unified pipeline.
-
Prompt-Unknown Promotion Attacks against LLM-based Sequential Recommender Systems
PUDA enables effective promotion of unpopular target items in black-box LLM sequential recommenders by using evolutionary LLM refinement to infer hidden prompts, training a surrogate model, and combining adversarial text revision with surrogate-generated poisoning sequences.
-
SAGE: A Service Agent Graph-guided Evaluation Benchmark
SAGE is a new multi-agent benchmark that formalizes service SOPs as dynamic dialogue graphs to measure LLM agents on logical compliance and path coverage, uncovering an execution gap and empathy resilience across 27 models in 6 scenarios.
-
Skill-Conditioned Visual Geolocation for Vision-Language Models
GeoSkill lets vision-language models improve geolocation accuracy and reasoning by maintaining an evolving Skill-Graph that grows through autonomous analysis of successful and failed rollouts on web-scale image data.
-
ToolPRM: Fine-Grained Inference Scaling of Structured Outputs for Function Calling
ToolPRM provides fine-grained intra-call process supervision via a new dataset and reward model, outperforming outcome and coarse-grained alternatives on function-calling benchmarks.
-
Exploiting LLM Agent Supply Chains via Payload-less Skills
Semantic Compliance Hijacking lets attackers hijack LLM agents by disguising malicious instructions as compliance rules in skills, reaching up to 77.67% success on confidentiality breaches and 67.33% on RCE while evading all tested scanners.
-
AgentGR: Semantic-aware Agentic Group Decision-Making Simulator for Group Recommendation
AgentGR uses semantic-aware LLM agents to simulate group decision dynamics and improve group recommendation accuracy over traditional aggregation methods.
-
Video Understanding Reward Modeling: A Robust Benchmark and Performant Reward Models
Introduces VURB benchmark and VUP-35K dataset to train discriminative and generative video reward models that achieve SOTA performance on VURB and VideoRewardBench.
-
Stayin' Aligned Over Time: Towards Longitudinal Human-LLM Alignment via Contextual Reflection and Privacy-Preserving Behavioral Data
A methodological framework and browser system BITE for collecting evolving user preferences on LLM outputs through context-triggered reflections and privacy-preserving data over time.
-
LocalAlign: Enabling Generalizable Prompt Injection Defense via Generation of Near-Target Adversarial Examples for Alignment Training
LocalAlign generates near-target adversarial examples via prompting and applies margin-aware alignment training to enforce tighter boundaries against prompt injection attacks.
-
Beyond Semantic Relevance: Counterfactual Risk Minimization for Robust Retrieval-Augmented Generation
CoRM-RAG uses a cognitive perturbation protocol to simulate biases and trains an Evidence Critic to retrieve documents that support correct decisions even under adversarial query changes.
-
Ethics Testing: Proactive Identification of Generative AI System Harms
Ethics testing is introduced as a systematic approach to generate tests that identify software harms induced by unethical behavior in generative AI outputs.
-
Call-Chain-Aware LLM-Based Test Generation for Java Projects
CAT improves line coverage by 18% and branch coverage by 22% over prior LLM test generation methods by adding call-chain and dependency context from static analysis to prompts.
-
GeoMind: An Agentic Workflow for Lithology Classification with Reasoned Tool Invocation
GeoMind applies an agentic workflow with tool-augmented modules and process supervision to outperform static models on lithology classification from well logs while producing traceable decisions.
-
Infection-Reasoner: A Compact Vision-Language Model for Wound Infection Classification with Evidence-Grounded Clinical Reasoning
Infection-Reasoner, a 4B VLM, reaches 86.8% accuracy on wound infection classification while producing rationales rated mostly correct by experts, via GPT-5.1 distillation followed by reinforcement learning.
-
Beyond Semantic Similarity: A Component-Wise Evaluation Framework for Medical Question Answering Systems with Health Equity Implications
VB-Score shows three major LLMs have severe failures in medical entity recognition and factual consistency, with 13.8% lower performance on chronic conditions affecting older and minority groups, indicating condition-based algorithmic discrimination.
-
RoTRAG: Rule of Thumb Reasoning for Conversation Harm Detection with Retrieval-Augmented Generation
RoTRAG retrieves Rules of Thumb to ground LLM reasoning for harm detection and severity classification in multi-turn dialogues, reporting roughly 40% relative F1 gains and 8.4% lower distributional error on two safety benchmarks while cutting redundant retrieval.
-
Structured Safety Auditing for Balancing Code Correctness and Content Safety in LLM-Generated Code
Dual Reasoning with explicit safety audits improves the new SUDS metric by 1.32x to 3.42x over baselines on code generation benchmarks containing injected harmful keywords.
-
GRASP: Grounded CoT Reasoning with Dual-Stage Optimization for Multimodal Sarcasm Target Identification
GRASP improves multimodal sarcasm target identification by anchoring visual regions in grounded chain-of-thought reasoning and using dual-stage optimization on a new balanced dataset.
-
VC-Soup: Value-Consistency Guided Multi-Value Alignment for Large Language Models
VC-Soup uses a cosine-similarity consistency metric to filter data, trains value-consistent policies, and applies linear merging with Pareto filtering to improve multi-value LLM alignment trade-offs.
-
GroupGPT: A Token-efficient and Privacy-preserving Agentic Framework for Multi-User Chat Assistant
GroupGPT decouples intervention timing from response generation via edge-cloud collaboration for multi-user chats, scoring 4.72/5 on the new MUIR benchmark of 2500 segments while cutting token use by up to 3x and adding privacy sanitization.
-
Compounding Disadvantage: Auditing Intersectional Bias in LLM-Generated Explanations Across Indian and American STEM Education
LLMs generate lower-quality STEM explanations for marginalized student profiles in Indian and American contexts, with intersectional compounding producing gaps of up to 2.55 grade levels.
-
SciHorizon-GENE: Benchmarking LLM for Life Sciences Inference from Gene Knowledge to Functional Understanding
SciHorizon-GENE is a large-scale benchmark evaluating LLMs on gene-to-function inference across four perspectives, revealing heterogeneity and challenges in faithful, complete, literature-grounded outputs.
-
Generative Bid Shading in Real-Time Bidding Advertising
GBS replaces two-stage bid landscape modeling with an autoregressive generative model plus reward-aligned policy optimization to improve short- and long-term advertiser surplus in real-time bidding.
-
Tuning Language Models for Robust Prediction of Diverse User Behaviors
BehaviorLM applies progressive fine-tuning in two stages to let LLMs predict both frequent anchor and rare tail user behaviors more robustly on real-world datasets.
-
Cheap Expertise: Mapping and Challenging Industry Perspectives in the Expert Data Gig Economy
AI data firms view human expertise as an extractable, low-cost resource to feed AI systems while treating institutional expertise as something needing liberation or reform to fit this model.
-
Uni-HOI:A Unified framework for Learning the Joint distribution of Text and Human-Object Interaction
Uni-HOI learns the joint distribution of text, human motion, and object motion using LLMs and VQ-VAEs in a two-stage training process for multiple HOI tasks.
-
Breakdowns in Conversational AI: Interactional Failures in Emotionally and Ethically Sensitive Contexts
Mainstream conversational models show escalating affective misalignments and ethical guidance failures during staged emotional trajectories, organized into a taxonomy of interactional breakdowns.
-
TIGFlow-GRPO: Trajectory Forecasting via Interaction-Aware Flow Matching and Reward-Guided Optimization
TIGFlow-GRPO uses a Trajectory-Interaction-Graph in conditional flow matching plus Flow-GRPO optimization to produce more accurate, socially compliant, and physically feasible trajectory forecasts on ETH/UCY and SDD datasets.
-
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.
-
Reinforcing 3D Understanding in Point-VLMs via Geometric Reward Credit Assignment
Geometric Reward Credit Assignment disentangles rewards to geometric tokens and adds reprojection consistency to boost 3D keypoint accuracy from 0.64 to 0.93 and bounding box IoU to 0.686 on a ShapeNetCore benchmark while preserving 2D performance.
-
An Empirical Study of Perceptions of General LLMs and Multimodal LLMs on Hugging Face
Hugging Face discussions show that access barriers, output quality, and setup complexity are the main user concerns for both general and multimodal LLMs.
-
Brainrot: Deskilling and Addiction are Overlooked AI Risks
AI safety literature overlooks cognitive deskilling and addiction risks from generative AI despite public concern about them.
- BEAR: Towards Beam-Search-Aware Optimization for Recommendation with Large Language Models