Browse all research

Every Research Watch entry, newest first. Search, or narrow by topic, format, peer review, publisher and date.

1306 entries

Sort

The summaries are open to everyone. Sign in, free, to read the full entry.

Preprint
Not peer reviewed
Artificial Intelligence

A preprint reports SPLASH, a system that switches LLM attention layouts live during serving

Original title: SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving

The authors describe SPLASH, a serving system that switches between attention layouts mid-run and reports notable throughput gains over fixed layouts.

Chuan Liu et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
From Quantum Motion
Quantum Computing

Researchers report erasure-based QEC scheme for shuttled silicon spin qubits

Original title: Erasure conversion for singlet-triplet spin qubits enables high-performance shuttling-based quantum error correction

A preprint proposes using singlet-triplet spin qubits as erasure qubits to boost error correction thresholds in shuttling-based silicon quantum computers.

Added 5 Oct 2026 · Released Jan 2026

Preprint
Not peer reviewed
From Meta
Artificial Intelligence

Preprint proposes Meta-SecAlign to cut prompt-injection risk without losing agent utility

Original title: Meta-SecAlign: Training LLMs against Prompt Injection for Robust Agents

The authors report a fine-tuning method that defends large language models against prompt injection while largely preserving their usefulness on agentic tasks.

Added 5 Oct 2026

Preprint
Not peer reviewed
From Tencent Hunyuan
Artificial Intelligence

A preprint reports RL gains in LLMs trace to a low-dimensional activation geometry

Original title: Learning to Steer, Steering to See: Unveiling the Geometry of RLVR in Large Language Models via Trainable Vectors

The authors find reinforcement learning's effect on large language models can be captured by a few trainable activation vectors, and build a stabiliser on this idea.

Added 5 Oct 2026

Preprint
Not peer reviewed
Quantum Computing

Preprint reports modular method to amplify and certify quantum code distance

Original title: Ultra-high-distance quantum memories from amplified qLDPC codes

A preprint describes 'distance amplifiers', a modular way to build and certify higher-distance quantum memories from smaller quantum LDPC code blocks.

Zijian Liang et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
Artificial Intelligence

Researchers report sparse circuit repairs can fix LLM reasoning errors without broad side effects

Original title: RESCUE: Repairing Language Model Errors to Sparse Circuits via Reinforcement Learning

A preprint describes locating and retuning small LLM circuits to fix math and medical QA errors while preserving other skills.

Chuanpu Liu et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
Quantum Computing

Researchers propose a projector-based ansatz to cut CNOT gates in quantum chemistry simulations

Original title: Projector Quantum Variational Ansatz

A preprint reports a projector-based VQE ansatz that needs fewer CNOT gates than ADAPT-VQE in early tests.

Thomas Dumontier et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
Quantum Computing

A preprint reports entanglement-efficient schemes for splitting QFT circuits across QPUs

Original title: Towards an Optimally Distributed Quantum Fourier Transform Circuit

The authors present two partitioning schemes that minimise entanglement costs for distributing the quantum Fourier transform across QPUs, validated on hardware.

Zachary Vernec et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
Quantum Computing

A preprint reports a constant-overhead scheme for injecting and ejecting quantum code states

Original title: Constant-Overhead Injection into Quantum Codes

The study presents a quantum code family that can encode and decode physical qubits with constant space-time overhead under circuit-level noise.

Louis Golowich et al. · Added 5 Oct 2026

Preprint
Not peer reviewed
From Google Quantum AI
Quantum Computing

Researchers report disorder-free localization in a lattice gauge theory on a quantum processor

Original title: Observation of disorder-free localization using a (2+1)D lattice gauge theory on a quantum processor

A preprint describes quantum circuit experiments showing localization without disorder in a lattice gauge theory, plus a faster way to sample disorder.

Added 5 Oct 2026 · Released Jul 2025

Journal article
Peer reviewed
From ARQUE Systems
Quantum Computing

Study maps how spin-qubit shuttling errors affect surface-code thresholds

Original title: Performance of the spin qubit shuttling architecture for a surface code implementation

The study models how shuttling errors in a spin-qubit bus architecture affect surface-code error-correction thresholds and qubit overhead.

Added 5 Oct 2026

Journal article
Peer reviewed
From IBM Quantum
Quantum Computing

Researchers report constant-depth logical measurement for magic state preparation

Original title: Constant-depth magic state cultivation with Clifford measurements by gauging

A peer-reviewed study reports a gauging-based method that measures a logical Clifford operator in constant circuit depth, improving on existing cultivation schemes.

Added 4 Oct 2026

Journal article
Peer reviewed
Quantum Computing

Hole spin qubit shows microwave photons, not noise, limit its coherence

Original title: Coherence of a hole spin flopping-mode qubit in a circuit quantum electrodynamics environment

Researchers report that a silicon hole spin flopping-mode qubit's coherence is mainly limited by photon effects rather than typical spin noise sources.

Léo Noirot et al. · Added 4 Oct 2026 · Released May 2026

Journal article
Review status unknown
Quantum Computing

Researchers report titanium layer cuts loss in tantalum qubit resonators

Original title: Reducing TLS loss in tantalum CPW resonators using titanium sacrificial layers

A journal study reports that an ultrathin titanium sacrificial layer sharply lowers two-level-system loss in tantalum superconducting resonators.

Zachary Degnan et al. · Added 4 Oct 2026 · Released Jan 2026

Preprint
Not peer reviewed
From IonQ
Quantum Computing

Preprint proposes trapped-ion blueprint for fault-tolerant quantum computing using LDPC codes

Original title: Fault-Tolerant Quantum Computing with Trapped Ions: The Walking Cat Architecture

A preprint outlines a trapped-ion architecture using cat states and LDPC codes, reporting simulated designs for fault-tolerant quantum computing.

Added 4 Oct 2026 · Released Apr 2026

Preprint
Not peer reviewed
Artificial Intelligence

A preprint reports a reinforcement learning framework that blends retrieval with reasoning

Original title: UR$^2$: Unify RAG and Reasoning through Reinforcement Learning

UR2 trains language models to decide when to retrieve information, aiming to match larger commercial models on several tasks.

Weitao Li et al. · Added 4 Oct 2026

Journal article
Review status unknown
Quantum Computing

Multichannel integrated quantum memory enables random access and high-dimensional storage

Original title: Random access and high dimensional integrated quantum memory

Researchers report an 11-channel solid-state quantum memory chip that supports random access and stores high-dimensional photon states.

Zhong-Wen Ou et al. · Added 4 Oct 2026

Journal article
Review status unknown
Quantum Computing

Researchers propose chip design for tunable squeezed light and single photons

Original title: Periodically poled thin-film lithium niobate ring Mach Zehnder coupling interferometer as an efficient quantum source of light

A study proposes a thin-film lithium niobate device that researchers say can generate both squeezed light and high-purity single photons on one chip.

Mrinmoy Kundu et al. · Added 4 Oct 2026 · Released Aug 2024

Preprint
Not peer reviewed
Artificial Intelligence

Preprint proposes boosting diverse reward models to curb reward hacking in RLHF

Original title: RMB: Reward Model Boosting Mitigates Reward Hacking

A preprint reports that combining diverse reward models through a boosting method improves reward accuracy and reduces reward hacking in RLHF.

Jiabin Fan et al. · Added 4 Oct 2026

Preprint
Not peer reviewed
Artificial Intelligence

Researchers report models keep usefulness but lose safety under coded-text shifts

Original title: LLM Alignment–Utility Asymmetry under Semantic-Preserving Transformations

A preprint finds that once models learn a simple text-substitution code, they keep most task performance but become far more likely to comply with harmful requests.

Mohan Li et al. · Added 4 Oct 2026