ML news

Sign In

by Amazon Research Publication

Amazon Research PublicationFollow

Using Cross-Domain Signals for Cold-Start Audiobooks

Unified cross-domain semantic retrieval for cold-start audiobook discovery
...
...
Amazon Research PublicationFollow

Concentration beats coverage: Impact-aware expert review for AI-assisted carbon footprint estimation

Concentration beats coverage: Impact-aware expert review for AI-assisted carbon footprint estimation
...
...
Amazon Research PublicationFollow

Answering Incomplete Rule Questions with AI and Logic

The question is a region: A neurosymbolic retrofit for categorical question answering over rule engines
...
...
Amazon Research PublicationFollow

Finding Hidden Data Corruption Beyond Schema Checks

Beyond schema validation: Characterizing structurally valid data corruption in multi-source analytical pipelines
...
...
Amazon Research PublicationFollow

How AI Agents Keep Cloud Infrastructure in Sync

NSync: Automated cloud infrastructure-as-code reconciliation with AI agents
...
...
Amazon Research PublicationFollow

ReCodeAgent: A multi-agent workflow for language-agnostic translation and validation of large-scale repositories

ReCodeAgent: A multi-agent workflow for language-agnostic translation and validation of large-scale repositories
...
...
Amazon Research PublicationFollow

How Shared Search Keeps Recommendations Ready for New Items

Keeping the index open: The recommendation-side cost of shared search and recommendation
...
...
Amazon Research PublicationFollow

Teaching AI Which Tokens Matter for Product Detection

Not all tokens earn the reward: Contrastive credit assignment for production detection
...
...
Amazon Research PublicationFollow

When Does Text Improve Time Series Forecasting?

When does multimodality lead to better time series forecasting?
...
...
Amazon Research PublicationFollow

Evaluating Safety Across Multimodal AI Models

UNISAF-BENCH: Intention-grounded safety evaluation for unified multimodal generation models
...
...
Amazon Research PublicationFollow

How LLMs Learn Auditable Rules for Trustworthy Decisions

Trustworthy LLM-based rule learning for auditable agentic decision-making
...
...
Amazon Research PublicationFollow

Distilling AI Reasoning for Smarter Product Upgrades

Distill globally, adapt locally: Reasoning distillation and product-type test-time training for scalable trade-up recommendation
...
...
Amazon Research PublicationFollow

How AI Maps Road Closures from Images and Road Networks

Multimodal road closure mapping via cross-modal graph attention
...
...
Amazon Research PublicationFollow

How Agents Turn Experience into Reusable Skills

Evo-harness: Context-to-harness skill compilation for self-evolving agents
...
...
Amazon Research PublicationFollow

How TaxCE Builds and Evaluates Hierarchical Taxonomies

TaxCE: A framework for automated taxonomy construction and evaluation at scale
...
...
Amazon Research PublicationFollow

Evaluating Shopping Assistants with the PACE Framework

PACES HOP: Evaluating personalized, actionable, compositional, and evidence-grounded shopping assistants
...
...
Amazon Research PublicationFollow

How LLMs Learn New Languages: Syntax, Algorithms, and Bugs

What does post-training actually teach LLMs about code? Isolating syntax, algorithms, and implementation fidelity with a novel language
...
...
Amazon Research PublicationFollow

How LLMs Build Better Playlists at Scale

LLM-based playlist curation: A music reasoning approach at scale
...
...
Amazon Research PublicationFollow

How Web Agents Can Be Poisoned by What They See

Poison once, exploit forever: Environment-injected memory poisoning attacks on web agents
...
...
Amazon Research PublicationFollow

How Rubrics Shape LLM Judge Accuracy and Reliability

PReMISE: Policy rubrics as measurement specifications for LLM judges
...
...