Isolater - Feed

Ax Sai Varun Kodathala 7/8/2026

aiAuthZ: Off-Host, Identity-Bound Authorization for AI Agents

aiAuthZ security framework for identity-bound authorization in AI agents, evaluating LLM vulnerability to tool-call forgery attacks across 15 models.

Ax Muhammad Zain Amin, Kibele Sebnem Yildirim 7/8/2026

Self-Review Reinforcement Learning (SRRL) with Cross-Episode Memory and Policy Distillation

Self-Review Reinforcement Learning method for LLMs using cross-episode memory and policy distillation to handle sparse/delayed environmental feedback.

Ax Yibo Hu, Jiaming Qu 7/8/2026

Most LLM Conformity Needs No Speaker: Measuring the Speaker-Free Floor in Peer-Pressure Benchmarks

Study showing LLM conformity to peer responses is largely confounded by repeated answers independent of speaker presence in benchmark tasks.

Ax Haonan Huang 7/8/2026

The yes-no bias of large language models reflects answer order and wording, not shifts in moral judgment

Analysis of LLM yes-no bias in binary judgments, isolating effects of answer order and wording from moral judgment shifts using psychometric methods.

Ax Sadia Kamal, Arefa Patwary, Anthony Marchiafava, Atriya Sen, Sagnik Ray Choudhury 7/8/2026

Prompt Robustness Is Task-Dependent: Comparing Objective and Belief-Style Questions in LLM Evaluation

Evaluation study comparing prompt robustness between objective and subjective LLM tasks, showing different sensitivity patterns across model families.

Ax Gregor Kobsik, Tim Elsner, Leif Kobbelt 7/8/2026

Harnessing Generative Image Models for Training-Free Primitive Shape Abstraction

Training-free approach using generative image models for 3D primitive shape abstraction without task-specific fine-tuning.

Ax Abhash Shrestha, Subigya Gautam, Anu Sapkota, Sanju Tiwari, Tek Raj Chhetri 7/8/2026

Whose fairness? Structural concentration in AI bias research

Analysis of structural concentration in AI bias research community, examining whose fairness definitions and debiasing frameworks are produced.

Ax Archie Chaudhury 7/8/2026

ResonatorLM: Causal Resonant Field Mixing for Efficient Long-Context Language Modelin

ResonatorLM architecture for efficient long-context language modeling using causal resonant field mixing as alternative to transformers.

Ax Giulia Lanzillotta, Mandana Samiei, Doina Precup, Razvan Pascanu, Claire Vernade 7/8/2026

To Retain or to Adapt? Generalizing Continual Learning

Continual learning research challenging retention-centered assumptions and prioritizing real-time adaptation in non-stationary environments.

Ax Abu Tyeb Azad, Ishita Sur Apan, Fahim Ahmed, Sumaiya Karim Katha, Ezharuddin Jubaer, Armun Alam, Pranjal Kumar Nandi, Amin Ahsan Ali, Aman Chadha, Md Mofijul Islam, AKM Mahbubur Rahman 7/8/2026

BaFCo: A Document Understanding Benchmark for Complex Bangla Form Comprehension

BaFCo benchmark dataset for Bangla form comprehension using multimodal LLMs, addressing low-resource language document understanding gaps.

Ax Kenneth Benavides, Josh Fleischer, Danti Chen 7/8/2026

EvalLoop: A Methodology for Evaluation-Driven Iterative Improvement of Business AI Systems

EvalLoop methodology for evaluation-driven iterative improvement of LLM-based business systems, focusing on diagnosis and fixing rather than static model selection.

Ax Illia Dovhoshliubnyi, Nima Soroush, Ashkan Sami, Alexander Brownlee 7/8/2026

What Do AI Agents Actually Change? An Empirical Taxonomy of Mutation Patterns in Performance-Improving Pull Requests

Empirical taxonomy of code mutations in AI-generated pull requests for performance optimization, analyzing 33k agent PRs.

Ax Damian Hodel, Jevin West, Aylin Caliskan 7/8/2026

RPAM: A Principled Metric for Evaluating Associations in Language Models with High Predictive Validity in Downstream Outputs

RPAM: metric for measuring semantic associations and biases in language models with predictive validity for downstream task performance.

Ax Benjamin Minhao Chen, Zhiyu Li 7/8/2026

Beyond Accuracy: How Humans Evaluate Legally Correct but Socially Controversial Legal Advice from Machines

Survey study examining human trust in AI-generated legal advice versus human lawyers when advice is legally correct but socially controversial.

Ax Haz Sameen Shahgir, Yufei Li, Frank Shyu, Luke Simon, Sandeep Pandey, Xi Liu, Yue Dong 7/8/2026

SCOReD: Student-Aware CoT Optimization for Recommendation Distillation

SCOReD: chain-of-thought distillation optimization for recommendation systems using student-aware training to avoid verbose reasoning traces.

Ax Mohammadreza Rashidi 7/8/2026

The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities

Survey of execution-layer security research for AI coding agents covering isolation, access control, TOCTOU vulnerabilities, and protocol threats.

Ax Mohammadreza Rashidi 7/8/2026

Unicode TAG-Block Concealment of Tool-Metadata Payloads in the Model Context Protocol: An Approval-View Fidelity Gap Across Three Independent Server Implementations

Security vulnerability analysis of Model Context Protocol implementations showing tool metadata concealment attacks across AI coding agent servers.

Ax Minho Kim 7/8/2026

When Should LLMs Search? Counterfactual Supervision for Search Routing

Counterfactual supervision method to train LLMs when to use external search versus parametric knowledge for improved task performance.

Ax Lejian Zhang, Xueyan Tang, Jing Tang 7/8/2026

Data-dependent Evaluations for Budgeted Submodular Maximization

Data-dependent evaluation methods for budgeted submodular maximization algorithms in machine learning optimization problems.

Ax Guang Yang, Brian Siyuan Zheng, Victoria Ebert, Noah A. Smith 7/8/2026

LEGATO 2: Toward Multimodal Sheet Music Recognition and Understanding

Legato 2 pipeline for optical music recognition processing sheet music sequentially to extract symbolic notation and semantic knowledge.

Ax Chuhao Zhou, Liquan Wang, Shuxin Cao, Xiangyu Chen, Yuxuan Hu, Boyu Ma, Animesh Garg, Jianfei Yang 7/8/2026

FORGE: Towards Functional Tool-Use Generalization via Keypoint Trajectory Reasoning

FORGE framework enables robots to generalize tool use across novel objects via keypoint trajectory reasoning for functional transfer learning.

Ax Yake Wei, Yuan Wang, Fengyun Rao, Jing Lyu, Di Hu 7/8/2026

Segmentation before Answering: Pixel Grounding for MLLM Visual Reasoning

SegAnswer: multimodal LLM method using pixel-level segmentation before answering to improve visual reasoning accuracy in image understanding tasks.

Ax Fred Mutisya, Oscar Onyango, Sarah Sitati, Syokau Ilovi, Aeesha NJ Malik, Brenda W'mosi, Brian Makini, Jalemba Aluuvala, Josiah Onyango, Rachael Kanguha Mmene, Steven Wanyee 7/8/2026

Complementary Roles of Image Classification and Vessel Segmentation in AI-Based Screening for Retinopathy of Prematurity Plus Disease in a Kenyan Preterm Cohort

AI-based screening system combining image classification and vessel segmentation for detecting retinopathy of prematurity in infants.

Ax Mingchen Li, Meikang Qiu, Zifan Peng, Heng Fan, Song Fu, Junhua Ding, Yunhe Feng 7/8/2026

Beyond Refusal: A Same-Lineage Study of Aligned and Abliterated LLMs for Vulnerability Analysis

Vulnerability analysis of aligned LLMs comparing same-lineage models to isolate safety behavior from architectural differences in code review terminology.

Ax Zhiyuan Chen, Jing Hu, Junzhe Wang, Yueyang Huang, Xinyi Yang, Zhaoyang Wang, Feng Zhu 7/8/2026

AbICL: In-Context Learning for Antigen-Specific Antibody Affinity Ranking

In-context learning approach for ranking antibody candidates by binding affinity using contextual information from labeled antigen-specific comparisons.

Ax Sishun Liu, Sajal Halder, Ke Deng, Yan Wang, Xiuzhen Zhang 7/8/2026

Unsupervised Anomaly Detection of Information Operations Users via Behavioral and Language Patterns

Unsupervised anomaly detection method for identifying information operations users via behavioral and language patterns on social media.

Ax Pan Li, Kai Chen, Shuai Chang, Shengzhi Zhang, Peizhuo Lv, Jinwen He 7/8/2026

Differentially Private Natural Gradient Descent

Natural gradient descent variant for differentially private training that accounts for loss curvature to improve optimization efficiency under privacy constraints.

Ax Yihua Liu 7/8/2026

Think Before You Grid-Search: Floor-First Triage for LLM Serving

Analytical framework for LLM serving optimization using floor-first residual-driven triage to estimate resource bottlenecks before grid search.

Ax Suneeta Mall, Vladimir Nekrasov, Ashnil Kumar, Sajith Karunasena, Aiden Nibali, Alix Bird, Mateo Diaz Shine, Jarrel Seah 7/8/2026

Harrison.Rad 1.5 Technical Report: A radiology foundation model that can draft reports from images, priors and clinical context

Harrison.Rad 1.5 multimodal foundation model for automated radiology report generation from images, clinical history, and prior studies.

Ax Cemil-Andrei Dilmac, Florinel-Alin Croitoru, Radu Tudor Ionescu 7/8/2026

Few-Medoids: An Embarrassingly Simple Coreset Selection Method for Few-Shot Knowledge Distillation

Simple coreset selection method using medoids for efficient few-shot knowledge distillation that surpasses random baseline sample selection.

Ax Jean-Francois Bonbhel 7/8/2026

K-ABENA: K-Adaptive Backpropagation with Error-based N-exclusion Algorithm : (Compensated Loss-Based Sample Exclusion with Unbiased Gradient Estimation)

Selective gradient computation framework reducing training cost by excluding low-loss samples from backward pass with unbiased gradient estimation.

Ax Mingyang Song, Luxin Xu, Haoyu Sun, Minzhou Pan, Yu Cheng, Bo Li 7/8/2026

PolicyShiftGuard: Benchmarking and Improving Policy-Adaptive Image Guardrails

Benchmark and methods for policy-adaptive image safety guardrails that generalize to policy changes without retraining.

Ax Ryota Tanaka, Taku Hasegawa, Kyosuke Nishida 7/8/2026

CMDR: Contextual Multimodal Document Retrieval

Contextual multimodal document retrieval benchmark and methods that preserve textual/visual content while resolving multi-page aggregation queries.

Ax Chunmin Xia, Jakub Harbaczewski, Nikhil Dsilva, Julie Raulin, Dominic Schneider, Achim Autenrieth 7/8/2026

Agentic AI for IPoDWDM Network Lifecycle Automation: An MCP-Enabled Architecture

Distributed multi-MCP architecture for vendor-agnostic SDN-based automation and autonomous control of multi-layer IPoDWDM networks with E2E service lifecycle automation.

Ax Krittanon Kaewtawee, Petmongkon Pornpichitsuwan, Natchaya Temyingyong, Nutnicha Laplamoon, Wachiravit Modecrua, Krittin Pachtrachai, Touchapon Kraisingkorn 7/8/2026

InfluMatch: Frontier-Quality KOL Search at 4B-Model Cost

Cost-efficient influencer matching system using three-stage cascade of small open-weight models instead of frontier LLM prompting for semantic matching on Thai marketing criteria.

Ax Riccardo Terrenzi, Serkan Ayvaz 7/8/2026

Faithful or Findable? Evaluating LLM-Generated Metadata for RDF Dataset Search

Evaluation of LLM-generated metadata for RDF dataset search, comparing rewriting and agentic graph-based generation methods for retrieval effectiveness and faithfulness.

Ax Chunmin Xia, Jakub Harbaczewski, Nikhil Dsilva, Julie Raulin, Dominic Schneider, Achim Autenrieth 7/8/2026

MCP-Enabled Agentic AI for Autonomous IPoDWDM Network Lifecycle Automation

Agentic AI architecture using MCP protocol for autonomous control and lifecycle automation of multi-vendor IPoDWDM networks with closed-loop control validation.

Ax Daniel Shalam, Emanuel Ben Baruch, Avi Ben Cohen, Tal Remez 7/8/2026

Propose and Attend: Training-free MLLM Grounding Confidence via Multi-Token Localized Attention

Training-free method to improve grounding confidence in multimodal LLMs by detecting hallucinated spatial/temporal predictions using multi-token localized attention.

Ax Daryna Dementieva, Nikolay Babakov, Kathy H\"ammerl, Ilseyar Alimova, Jind\v{r}ich Libovick\'y, Shu Okabe, Miras Baisbay, Lukas Edman, Abrorkhon Inomkhujaev, Antonia Karamolegkou, Mateusz Lango, Volkan \"Ozer, Nikola Selic, Subhankar Swain, Tsedeniya Kinfe Temesgen, Galit Bary Weisberg, Alexander Fraser 7/8/2026