4/17/2026
Enabling Agents to Communicate Entirely in Latent Space
Interlat framework enables LLM-based agents to communicate directly in latent space bypassing tokenization, improving information depth for collaborative problem-solving.
4/17/2026
Towards Deploying VLA without Fine-Tuning: Plug-and-Play Inference-Time VLA Policy Steering via Embodied Evolutionary Diffusion
VLA-Pilot enables inference-time steering of vision-language-action models for robotic manipulation without fine-tuning via embodied evolutionary diffusion.
4/17/2026
DocVAL: Validated Chain-of-Thought Distillation for Grounded Document VQA
DocVAL distills validated chain-of-thought reasoning from large VLMs to compact models for document VQA with precise spatial grounding and low latency.
4/17/2026
Enhancing Large Language Model-Based Systems for End-to-End Circuit Analysis Problem Solving
End-to-end circuit analysis system leveraging multimodal LLMs for diagram interpretation and mathematical reasoning in engineering problem solving.
4/17/2026
Learning to Plan, Planning to Learn: Adaptive Hierarchical RL-MPC for Sample-Efficient Decision Making
Hierarchical RL-MPC fusion approach for sample-efficient planning problems that couples reinforcement learning with model predictive control.
4/17/2026
Enhancing LLM-Based Neural Network Generation: Few-Shot Prompting and Efficient Validation for Automated Architecture Design
Few-shot prompting approach using LLMs for automated neural network architecture design in computer vision with efficient validation methods.
4/17/2026
RLPO: Residual Listwise Preference Optimization for Long-Context Review Ranking
RLPO applies listwise preference optimization to long-context e-commerce review ranking, improving calibration of top-k rankings over pointwise approaches.
4/17/2026
Adaptive Layer Selection for Layer-Wise Token Pruning in LLM Inference
Adaptive layer selection method for layer-wise token pruning in LLM inference that learns which layers benefit most from KV cache reduction.
4/17/2026
ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning
ORBIT enables controllable multi-budget reasoning in Large Reasoning Models by dynamically balancing exploration-exploitation for appropriate inference-time computation allocation.
4/17/2026
TopoDIM: One-shot Topology Generation of Diverse Interaction Modes for Multi-Agent Systems
TopoDIM framework optimizes communication topology in LLM-based multi-agent systems using one-shot diverse interaction mode generation for reduced latency and computation.
4/17/2026
LLMOrbit: A Circular Taxonomy of Large Language Models -From Scaling Walls to Agentic AI Systems
LLMOrbit comprehensive circular taxonomy surveying 50+ LLMs from 2019-2025 across architectural innovations, scaling approaches, and agentic AI system evolution.
4/17/2026
HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding
HERMES uses hierarchical KV cache memory for efficient streaming video understanding in multimodal LLMs without training, maintaining real-time performance and low GPU overhead.
4/17/2026
Multi-Persona Thinking for Bias Mitigation in Large Language Models
Multi-Persona Thinking framework reduces social bias in LLMs through inference-time multi-perspective reasoning guidance with contrasting viewpoints.
4/17/2026
Rethinking LLM-Driven Heuristic Design: Generating Efficient and Specialized Solvers via Dynamics-Aware Optimization
LLM-driven heuristic design framework that uses LLMs to generate and refine combinatorial optimization solvers with dynamics-aware evaluation beyond endpoint-only metrics.
4/17/2026
POP: Prefill-Only Pruning for Efficient Large Model Inference
POP proposes stage-aware structured pruning for LLMs/VLMs that differentiates between prefill and decode stages to improve inference efficiency while maintaining accuracy.
4/17/2026
Attack Selection Reduces Safety in Concentrated AI Control Settings against Trusted Monitoring
Study of attack selection vulnerability in monitored AI systems, analyzing how AI can evade detection by selecting which attacks to execute.
4/17/2026
IROSA: Interactive Robot Skill Adaptation using Natural Language
Framework combining foundation models with imitation learning for open-vocabulary robot skill adaptation from limited data for industrial deployment.
4/17/2026
MM-tau-p$^2$: Persona-Adaptive Prompting for Robust Multi-Modal Agent Evaluation in Dual-Control Settings
Evaluation framework for persona-adaptive LLM agents in dual-control settings with multimodal capabilities.
4/17/2026
Prompt Injection as Role Confusion
Analyzes prompt injection attacks as role confusion where models infer text source from style not origin.
4/17/2026
Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation
CAP-TTA test-time adaptation framework for debiasing LLMs on out-of-distribution high-bias prompts using LoRA.
4/17/2026
V-Reflection: Transforming MLLMs from Passive Observers to Active Interrogators
V-Reflection enables MLLMs to actively interrogate visual inputs and reduce hallucinations via visual reasoning.
4/17/2026
XMark: Reliable Multi-Bit Watermarking for LLM-Generated Texts
Multi-bit watermarking method for LLM-generated text enabling attribution and tracing with improved efficiency.
4/17/2026
VideoStir: Understanding Long Videos via Spatio-Temporally Structured and Intent-Aware RAG
VideoStir uses spatio-temporally structured RAG to scale MLLMs to long videos while preserving structure.
4/17/2026
AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent
Technical report on client-side optimization for LLM-based agents deployed in real-world applications.
4/17/2026
Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning
Unified framework connecting post-training methods for LLMs including SFT, preference optimization, RL, and process supervision.
4/17/2026
Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model
Studies inference-time safety improvements in LLMs via deliberative alignment and attribution of unsafe behavior to base models.
4/17/2026
Critical-CoT: A Robust Defense Framework against Reasoning-Level Backdoor Attacks in Large Language Models
Defense framework (Critical-CoT) against reasoning-level backdoor attacks in LLMs using chain-of-thought robustness.
4/17/2026
You Only Judge Once: Multi-response Reward Modeling in a Single Forward Pass
Discriminative reward model scoring multiple LLM responses in single forward pass for efficient comparative reasoning.
4/17/2026
METRO: Towards Strategy Induction from Expert Dialogue Transcripts for Non-collaborative Dialogues
Method to induce dialogue strategies from expert transcripts using LLMs for non-collaborative dialogue agents.
4/17/2026
METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models
Benchmark (METER) evaluating causal reasoning capabilities in LLMs across three levels of causal hierarchy.
4/17/2026
Not All Forgetting Is Equal: Architecture-Dependent Retention Dynamics in Fine-Tuned Image Classifiers
Studies forgetting dynamics in fine-tuned image classifiers across architectures to improve curriculum design and data pruning.
4/17/2026
I made an 80B local model ship a 295-test RAG codebas
Multi-agent framework (C.E.H.) autonomously built and tested a 295-test RAG codebase using an 80B local model. Open-source implementation.
4/17/2026
Poisoning AI Training Data
Technical analysis demonstrating how easy it is to poison AI training data via fake web content. Includes practical attack example.
4/17/2026
Solving Physics Olympiad via reinforcement learning on physics simulators
Research paper on reinforcement learning approach (Sim2Reason) for solving Physics Olympiad problems using physics simulators. Code and paper available.
4/17/2026
Aurora
Aurora: research on speculative decoding for LLM inference with draft model staleness solutions. 40+ models benchmarked. Paper and code available.
4/17/2026
Software Is About to Get Cheap
Opinion piece on implications of AI making software development 10x cheaper and faster. Discussion of SaaS business model impact.
4/17/2026
AST and LLM Navigation Tool
Developer tool using AST and LLM to navigate large codebases by intent. Creates navigable summaries for faster code exploration.
4/17/2026
LLM inference infrastructure for a systems audience
Systems-level technical overview of LLM inference infrastructure, serving runtimes, and ecosystem from infrastructure developer perspective.
4/17/2026
The first signed, drift-monitored W3C WebMCP manifest
W3C WebMCP standard enables AI agents to discover business tools via signed manifests, allowing agents like ChatGPT/Claude to complete tasks like booking services.
4/17/2026
Parcae: Doing more with fewer parameters using stable looped models
Parcae: stable looped language model architecture achieving Transformer-scale quality with fewer parameters and predictable training.
4/17/2026
I built 7 AI agents that attack the same task in parallel – armyai.app
Platform with 7 parallel AI agents executing specialized tasks simultaneously on complex problems using Groq and Cerebras.
4/17/2026
Autoresearch on Steroids with Sandboxes
Using sandboxes for LLM agents to safely execute autoresearch: agents propose code changes, run them, measure metrics iteratively.
4/17/2026
Google Opens Personal Intelligence AI to All US Users
Google launches Personal Intelligence feature to all US users, enabling Gemini to access Gmail and Photos for personalized responses.
4/17/2026
Anyone used this new tool?
Vybsly Search API provides AI agents and models actual web page content instead of links for real-time information access.
4/17/2026
Server Nexe – Modular, local AI server with persistent memory, zero cloud
Server Nexe: local AI server with persistent memory, no cloud dependency. Open source project for running AI locally with full control.
4/17/2026
Show HN: Swarm – Get consistent results from Claude Code
Swarm framework for tuning Claude Code agent prompts and memory to achieve consistent multi-agent results. Open source tool for developers.
4/17/2026
Weft: A Centroid Language for LLMs
Weft: a language ChatGPT invented that other frontier LLMs can understand zero-shot without training. Exploration of inter-LLM communication and model coherence.
4/16/2026
Connect maker devices to Claude Code and Cowork
Anthropic Claude API enabling BLE hardware integration with Claude Code and Cowork. LLM developer tool for building hardware experiences with Claude.
4/16/2026
Show HN: 70% → 100% LLM accuracy by changing the representation, not the model
Technique improving LLM accuracy on math/code tasks from 70% to 100% by representing numbers as spatial matrices instead of text. Novel representation approach for LLM reasoning.
4/16/2026
Isolating AI Coding Agents on Bare Metal
Case study on isolating autonomous AI coding agents on bare metal for healthcare and finance projects. Production AI agent infrastructure, runtime isolation, and deployment patterns.