Natural Language Processing

grammarlyJan 28, 2026Original article

AI Assistants vs. AI Agents: What’s the Difference and When to Use Each (opens in new tab)

While AI assistants and agents often share the same large language model foundations, they serve distinct roles based on their level of autonomy and task complexity. Assistants operate on a reactive "prompt-response" loop for immediate, single-step tasks, whereas agents function as semi-independent systems capable of planning and executing multistep workflows to achieve a broader goal. Ultimately, the most effective AI strategy involves leveraging assistants for quick, guided interactions while utilizing agents to manage complex, coordinated projects that require memory and tool integration. ### Reactive vs. Proactive AI Architectures * Assistants are reactive tools that follow a "prompt-response" loop, similar to a tennis match where the user must always serve to initiate action. * Agents are proactive and semi-independent; once given a high-level goal, they can decompose it into actionable steps and execute them with minimal step-by-step direction. * In a practical scenario, an assistant might summarize meeting notes upon request, whereas an agent can organize those notes, assign tasks in a project management tool, and schedule follow-ups automatically. ### Technical Capabilities and Coordination * Both tools utilize Large Language Models (LLMs) to understand natural language, but agents incorporate advanced features like long-term memory and cross-app integrations. * Memory allows agents to retain feedback and results from previous interactions to deliver better outcomes over time, while integrations enable them to act on the user's behalf across different software platforms. * The two systems often work in tandem: the assistant acts as the front-facing interface (the "waiter") for user commands, while the agent acts as the back-end engine (the "kitchen") that performs the orchestration. ### Balancing Control and Complexity * AI assistants provide high user control and instant setup, making them ideal for "out of the box" tasks like grammar checks, rephrasing text, or answering quick questions. * AI agents excel at reducing cognitive load by managing "moving parts" like deadline tracking, organizing inputs from different stakeholders, and maintaining project states across various tools. * Grammarly’s implementation of agents serves as a technical example, moving beyond simple text revision to offer context-aware suggestions that help with brainstorming, knowledge retrieval, and predicting audience reactions. To maximize productivity, users should delegate isolated, high-control tasks to AI assistants while allowing AI agents to handle the background orchestration of complex projects. Success with these tools depends on maintaining human oversight, using assistant-led prompts to provide the regular feedback that agents need to refine their autonomous workflows.

Natural Language Processing Large Language Models AI Agents Automation+2

netflixJan 27, 20263 min readCurated summary

The AI Evolution of Graph Search at Netflix

Netflix is evolving Graph Search from structured DSL queries toward natural-language search using large language models (LLMs). The goal is to let users describe what they want in everyday language while preserving the accuracy and trustworthiness required by Netflix’s complex, federated GraphQL data. Rather than replacing existing applications, Netflix plans to augment them with AI-generated filters and future retrieval-augmented generation capabilities. ## Why Natural-Language Search Is Needed - Graph Search currently relies on a Filter DSL, with applications translating UI interactions into structured queries. - Netflix has hundreds of applications with inconsistent query-building experiences, forcing users to learn different interfaces. - Some indexes contain hundreds of filterable fields, making large forms slow and cumbersome even for subject-matter experts. - Users naturally express goals in language such as “show all movies from the 90s about robots from the US,” not through query builders or DSL syntax. - Natural-language input could reduce friction while allowing each application to retain its own domain-specific presentation. ## Converting Text into Graph Search Filters - The core task is translating an ambiguous natural-language request into a valid Graph Search Filter DSL statement. - Graph Search indexes are defined through GraphQL queries containing typed fields, including booleans, strings, enums, and controlled vocabularies. - Generated filters can combine: - Comparisons such as `>` and `==` - Inclusion or exclusion operators such as `IN` - Logical operators such as `AND` - Netflix evaluates generated queries at three levels: - **Syntactic correctness:** The statement follows the DSL grammar and can be parsed. - **Semantic correctness:** The query uses existing fields, respects field types, and selects valid controlled-vocabulary values. - **Pragmatic correctness:** The filter accurately reflects the user’s intended meaning. ## Context Engineering for the LLM - The LLM needs index metadata to generate semantically valid filters. - Netflix derives much of this context from GraphQL schemas, including: - Field paths - Field descriptions from schema comments - Field types - Valid enum or controlled-vocabulary values - Controlled vocabularies define finite, governed sets of values, such as countries, and prevent generated queries from using invalid alternatives. - Supplying all metadata works for simple examples but does not scale: - Some indexes contain hundreds of fields. - Some vocabularies contain thousands of values. - Larger prompts increase latency and can reduce generation accuracy. - Netflix therefore needs ways to provide the LLM with relevant metadata without overwhelming its context, while still grounding generated queries in the actual schema and allowed values. Netflix’s approach combines schema-aware context, LLM-based query generation, and validation to make natural-language Graph Search practical. The key recommendation is to use AI as an augmentation layer over existing Graph Search applications, with strong grounding and correctness checks rather than treating generated queries as inherently reliable.

Read original

Natural Language Processing Large Language Models RAG GraphQL+4

googleJan 22, 20263 min readCurated summary

Small models, big results: Achieving superior intent extraction through decomposition

Small multimodal models can outperform much larger models at extracting user intent from UI interaction trajectories when the task is decomposed. Google’s approach first summarizes each screen and interaction, then derives an overall intent from those summaries. This enables accurate, faster, and more privacy-preserving on-device understanding without sending sensitive UI data to servers. ## Why On-Device Intent Understanding Matters - Understanding what users are doing across mobile and web interfaces can help agents anticipate useful next actions. - Large multimodal models perform well but often require server-side processing, introducing latency, cost, and privacy risks. - The goal is to make intent understanding practical with smaller models running directly on devices. ## Two-Stage Intent Extraction ### Screen and Interaction Summaries For each interaction, a small multimodal model examines a sliding window of three screens: the previous, current, and next screens. It generates information about: - Salient context on the current screen. - Actions the user just performed. - A speculation about what the user is trying to accomplish. This converts raw screenshots and actions into structured, manageable event summaries. ### Intent Extraction from Summaries A fine-tuned small model then processes the sequence of summaries and produces a single concise intent statement. The authors improve this stage through: - **Fine-tuning:** Training on examples of high-quality intent statements helps the model retain relevant details and discard noise. - **Label preparation:** Training intents are stripped of details absent from the summaries, reducing hallucinated information. - **Removing speculation:** Speculative fields help create richer individual summaries but are excluded from the second stage because they can confuse intent extraction. ## Evaluation with Atomic Facts - The authors use the Bi-Fact evaluation method to compare predicted intents with reference intents. - Each intent is split into indivisible “atomic facts,” such as “a one-way flight” or the separate origin and destination in a flight request. - The method measures: - **Recall:** How many reference facts were captured. - **Precision:** How many predicted facts are supported by the reference. - **F1:** The balance between precision and recall. - Tracking facts through both stages also reveals where details are lost or hallucinated. ## Results - The decomposed method outperformed chain-of-thought prompting and end-to-end fine-tuning. - Improvements held across both mobile and web interaction trajectories. - Results were consistent across Gemini and Qwen2 base models. - Gemini 1.5 Flash 8B achieved results comparable to Gemini 1.5 Pro while offering substantially lower cost and faster processing. - On mobile data, the small-model approach approached the performance of the larger Gemini Pro model. The study suggests that decomposing intent understanding into local summarization followed by sequence-level extraction is an effective path toward accurate, private, and efficient on-device assistants. As mobile hardware and small models improve, this technique could support a broad range of assistive features.

Read original

Natural Language Processing Gemini Multimodal AI Fine Tuning+4

kakaoDec 21, 2025Original article

Development of an Ultra-lightweight Classic (opens in new tab)

Kakao developed a specialized, lightweight morphological analyzer to meet the strict resource constraints of mobile environments where modern deep-learning models are often too heavy. By opting for a classical Viterbi-based approach implemented in C++20, the team successfully reduced the library's binary size to approximately 200KB while ensuring high performance. This development highlights how traditional algorithmic optimization and careful language selection remain vital for mobile software efficiency. ## The Choice of C++ over Rust - While Rust was considered for its safety, it was ultimately rejected because its default binary size (even with optimization) reached several megabytes, which was too large for the specific project requirements. - C++ was chosen because mobile platforms like iOS and Android already include standard libraries (libc++ or libstdc++), allowing the final analyzer binary to be stripped down to core logic. - The project utilized C++20 features such as Concepts and `std::span` to replace older patterns like SFINAE and `gsl::span`, resulting in more readable and maintainable code without sacrificing performance. ## Trie Compression using LOUDS - To minimize the dictionary size, the team implemented a LOUDS (Level-Order Unary Degree Sequence) structure, which represents a Trie using a bit sequence instead of pointers. - This approach provides a compression rate near the information-theoretic lower bound, allowing approximately 760,000 nodes to be stored in just 9.4MB. - Further optimization was achieved through a custom encoding scheme that represents Hangul in 2 bytes and English in 1 byte, significantly reducing the dictionary's memory footprint compared to standard UTF-8. ## Optimizing the Select Bit Operation - Initial performance profiling showed that the `select0` operation (finding the N-th zero in a bit sequence) consumed 90% of the dictionary search time due to linear search overhead. - The solution involved dividing the bit sequence into 64-bit chunks and storing the cumulative count of zeros at each chunk boundary in a separate array. - By using binary search to find the correct chunk and applying parallel bit-counting techniques for intra-chunk searching, the dictionary search time was reduced from 165ms to 10ms. - These optimizations led to a total analysis time improvement from 182ms to 28ms, making the tool highly responsive for real-time mobile use. For mobile developers facing strict hardware limitations, this project proves that combining classical data structures like LOUDS with modern low-level language features can yield performance and size benefits that deep learning alternatives currently cannot match.

Natural Language Processing AI Mobile Development Optimization+5

googleDec 14, 2025Original article

Gemini provides automated feedback for theoretical computer scientists at STOC 2026 (opens in new tab)

Google Research launched an experimental program for the STOC 2026 conference using a specialized Gemini model to provide automated, rigorous feedback on theoretical computer science submissions. By identifying critical logical errors and proof gaps within a 24-hour window, the tool demonstrated that advanced AI can serve as a powerful pre-vetting collaborator for high-level mathematical research. The overwhelmingly positive reception from authors indicates that AI can effectively augment the human peer-review process by improving paper quality before formal submission. ## Advanced Reasoning via Inference Scaling - The tool utilized an advanced version of Gemini 2.5 Deep Think specifically optimized for mathematical rigor. - It employed inference scaling methods, allowing the model to explore and combine multiple possible solutions and reasoning traces simultaneously. - This non-linear approach to problem-solving helps the model focus on the most salient technical issues while significantly reducing the likelihood of hallucinations. ## Structured Technical Feedback - Feedback was delivered in a structured format that included a high-level summary of the paper's core contributions. - The model provided a detailed analysis of potential mistakes, specifically targeting errors within lemmas, theorems, and logical proofs. - Authors also received a categorized list of minor corrections, such as inconsistent variable naming and typographical errors. ## Identified Technical Issues and Impact - The pilot saw high engagement, with over 80% of STOC 2026 submitters opting in for the AI-generated review. - The tool successfully identified "critical bugs" and calculation errors that had previously evaded human authors for months. - Survey results showed that 97% of participants found the feedback helpful, and 81% reported that the tool improved the overall clarity and readability of their work. ## Expert Verification and Hallucinations - Because the users were domain experts, they were able to act as a filter, distinguishing between deep technical insights and occasional model hallucinations. - While the model sometimes struggled to parse complex notation or interpret figures, authors valued the "neutral tone" and the speed of the two-day turnaround. - The feedback was used as a starting point for human verification, allowing researchers to refine their arguments rather than blindly following the model's output. ## Future Outlook and Educational Potential - Beyond professional research, 75% of surveyed authors see significant educational value in using the tool to train students in mathematical rigor. - The experiment's success has led to 88% of participants expressing interest in having continuous access to such a tool throughout their entire research and drafting process. The success of the STOC 2026 pilot suggests that researchers should consider integrating specialized LLMs early in the drafting phase to catch "embarrassing" or logic-breaking errors. While the human expert remains the final arbiter of truth, these tools provide a necessary layer of automated verification that can accelerate the pace of scientific discovery.

Natural Language Processing AI Large Language Models Generative AI+5

pinterestDec 10, 20253 min readCurated summary

LLM-Powered Relevance Assessment for Pinterest Search

Pinterest Search uses fine-tuned multilingual LLMs to assess search-result relevance at a much larger scale than human labeling allows. The approach combines five-level relevance classification, stratified query sampling, and paired A/B-test evaluation to detect smaller overall effects and differences across query types. XLM-RoBERTa-large provides a practical balance of accuracy and cost, achieving strong agreement with human judgments while enabling substantially faster labeling. ## Relevance Measurement Challenges - Search relevance measures how well Pins satisfy a user’s query, rather than merely reflecting past engagement. - Human annotations are expensive and limited in volume. - Previous sampling designs could detect only relatively large topline changes, with minimum detectable effects (MDEs) around 1.3%–1.5%. - Limited labels also made it difficult to measure heterogeneous effects across query interests or popularity segments. ## Fine-Tuned LLM Relevance Model - Pinterest defines relevance using five labels: - L5: Highly Relevant - L4: Relevant - L3: Marginally Relevant - L2: Irrelevant - L1: Highly Irrelevant - A cross-encoder model predicts the relevance of each Pin for a query. - Open-source multilingual models are fine-tuned on human-annotated examples using multiclass cross-entropy loss. - Pin representations include: - Titles and descriptions - BLIP-generated image captions - Linked-page titles and descriptions - Board titles where Pins were saved - Highly engaged query tokens associated with the Pin - Models tested included multilingual BERT, T5, mDeBERTa, XLM-RoBERTa, and Llama 3. - The final relevance label is selected from the model’s five output scores using argmax. ## Stratified Query Sampling - Lower LLM labeling costs allow Pinterest to use much larger and more detailed samples. - Queries are stratified using: - A DistilBERT-based query-to-interest model - Query popularity, based on how many users issue each query - Stratification improves representativeness and reduces variance by grouping similar queries. - Pinterest moved from simple random sampling to stratified sampling with optimal allocation across strata. - Most of the MDE improvement came from variance reduction through stratification. - The redesigned process reduced MDEs from approximately 1.3%–1.5% to 0.25% or less. ## LLM-Based A/B-Test Measurement - Pinterest samples paired queries from control and treatment groups. - Pairing controls for differences between queries, which are a major source of relevance variance. - For each query, the top 25 results are retained and labeled by the LLM. - Query-level relevance is measured using sDCG@25, a variant of nDCG that assumes an unlimited supply of highly relevant L5 results. - Results are aggregated into topline experiment metrics. - Heterogeneous effects are analyzed by query popularity and interest categories such as beauty, fashion, and art. - The Benjamini–Hochberg procedure controls the false discovery rate when testing multiple segments. ## Model Choice and Validation - XLM-RoBERTa-large was selected for its balance of quality and efficiency. - On a single A10G GPU, it can label 150,000 rows in about 30 minutes. - Llama 3–8B produced slightly better accuracy but required roughly six times the inference time and cost. - LLM labels matched human labels exactly for 73.7% of Pins. - A total of 91.7% of predictions differed from human ratings by no more than one relevance point. Pinterest’s approach makes relevance evaluation cheaper, faster, and more statistically sensitive. Fine-tuned LLMs paired with stratified sampling are recommended for search experimentation when human labeling cannot provide enough coverage to detect small or heterogeneous ranking effects.

Read original

Natural Language Processing Large Language Models Fine Tuning Ab Testing+4

awsDec 2, 2025Original article

Amazon Bedrock AgentCore adds quality evaluations and policy controls for deploying trusted AI agents (opens in new tab)

AWS has introduced several new capabilities to Amazon Bedrock AgentCore designed to remove the trust and quality barriers that often prevent AI agents from moving into production environments. These updates, which include granular policy controls and sophisticated evaluation tools, allow developers to implement strict operational boundaries and monitor real-world performance at scale. By balancing agent autonomy with centralized verification, AgentCore provides a secure framework for deploying highly capable agents across enterprise workflows. **Governance through Policy in AgentCore** * This feature establishes clear boundaries for agent actions by intercepting tool calls via the AgentCore Gateway before they are executed. * By operating outside of the agent’s internal reasoning loop, the policy layer acts as an independent verification system that treats the agent as an autonomous actor requiring permission. * Developers can define fine-grained permissions to ensure agents do not access sensitive data inappropriately or take unauthorized actions within external systems. **Quality Monitoring with AgentCore Evaluations** * The new evaluation framework allows teams to monitor the quality of AI agents based on actual behavior rather than theoretical simulations. * Built-in evaluators provide standardized metrics for critical dimensions such as helpfulness and correctness. * Organizations can also implement custom evaluators to ensure agents meet specific business-logic requirements and industry-specific compliance standards. **Enhanced Memory and Communication Features** * New episodic functionality in AgentCore Memory introduces a long-term strategy that allows agents to learn from past experiences and apply successful solutions to similar future tasks. * Bidirectional streaming in the AgentCore Runtime supports the deployment of advanced voice agents capable of handling natural, simultaneous conversation flows. * These enhancements focus on improving consistency and user experience, enabling agents to handle complex, multi-turn interactions with higher reliability. **Real-World Application and Performance** * The AgentCore SDK has seen rapid adoption with over 2 million downloads, supporting diverse use cases from content generation at the PGA TOUR to financial data analysis at Workday. * Case studies highlight significant operational gains, such as a 1,000 percent increase in content writing speed and a 50 percent reduction in problem resolution time through improved observability. * The platform emphasizes 100 percent traceability of agent decisions, which is critical for organizations transitioning from reactive to proactive AI-driven operations. To successfully scale AI agents, organizations should transition from simple prompt engineering to a robust agentic architecture. Leveraging these new policy and evaluation tools will allow development teams to maintain the necessary control and visibility required for customer-facing and mission-critical deployments.

Natural Language Processing AI AI Agents Generative AI+5

dropboxOct 2, 20253 min readCurated summary

A practical blueprint for evaluating conversational AI at scale

Conversational AI systems depend on many probabilistic stages, so even small changes can cause unexpected regressions. Dropbox Dash’s experience shows that evaluation should be treated like production engineering: systematic, repeatable, and required before changes are shipped. The approach combines curated datasets, actionable metrics, LLM-based judging, and human review. ## Evaluation as a Development Discipline - AI pipelines include intent classification, retrieval, ranking, prompt construction, inference, and safety filtering. - Changes to any stage can affect final answer quality in unpredictable ways. - Dropbox initially used ad-hoc testing, but shifted to a standardized process in which every model, prompt, or retrieval change had to pass evaluation before merging. - The evaluation framework covers datasets, metrics, tools, and workflows. - Future-proof evaluation must extend beyond text to images, video, and audio. ## Curating Public and Internal Datasets - Public datasets established baseline retrieval and question-answering performance: - **Natural Questions** tested retrieval from very large documents. - **MS MARCO** emphasized queries requiring multiple document hits. - **MuSiQue** tested multi-hop reasoning. - Internal datasets captured real-world usage from anonymized Dropbox employee queries and content. - Representative query sets reflected actual user behavior, using proxy labels or internal annotators. - Representative content sets focused on shared files, documentation, and connected data sources. - LLM-generated synthetic questions and answers covered tables, images, tutorials, and factual lookups. - These datasets became the foundation for automated pass/fail checks and experiment monitoring. ## Why Traditional Metrics Fall Short - Metrics such as BLEU, ROUGE, METEOR, BERTScore, and embedding similarity are fast and useful for detecting major regressions. - They often fail to measure production requirements, including: - Correct source citations - Factual accuracy - Valid file names and references - Reliable table parsing - Clear formatting - High ROUGE or BERTScore can coexist with hallucinations, missing citations, or buried factual errors. - Production evaluation therefore requires metrics tied directly to task requirements. ## Using LLMs as Evaluation Judges - LLM judges can assess dimensions traditional metrics miss, including: - Whether an answer addresses the query - Whether claims are supported by context - Citation correctness - Clarity, formatting, and tone - Judges receive the query, model answer, source context, and sometimes a hidden reference answer. - They return structured scores—scalar or categorical—alongside explanations. - Rubrics and judge models must themselves be tested, calibrated, versioned, and improved. - Specialized judges may be needed for particular languages or technical domains. ## Calibration and Human Review - Periodic manual labeling of sampled outputs created calibration sets for comparing human and judge-model decisions. - The team tracked agreement and judge drift over time. - Engineers manually reviewed 5–10% of each release’s regression suite. - Disagreements were investigated as either judge-prompt defects or model errors. - Recurring problems led to prompt revisions or more detailed scoring criteria. Dropbox’s evaluation-first approach treats AI changes like code changes: test them against realistic datasets, score them with task-specific rubrics, validate the evaluators, and retain human oversight. This makes conversational AI development more reliable as systems and modalities grow more complex.

Read original

Natural Language Processing Large Language Models Prompt Engineering RAG+4

googleSep 23, 2025Original article

AfriMed-QA: Benchmarking large language models for global health (opens in new tab)

AfriMed-QA is a comprehensive benchmarking suite designed to address the critical gap in medical LLM evaluation for African healthcare contexts. Developed through a partnership between Google Research and a pan-African consortium, the project demonstrates that current models often struggle with geographic distribution shifts in disease and localized linguistic nuances. The researchers conclude that diverse, region-specific datasets are essential for training equitable AI tools that can safely provide clinical decision support in low-resource settings. ## Limitations of Western-Centric Benchmarks * Existing medical benchmarks like USMLE MedQA focus on Western clinical contexts, which may not generalize to other regions. * Models trained on traditional datasets often fail to account for specific distribution shifts in disease types and cultural symptom descriptions. * The lack of diverse data makes it difficult to assess how LLMs handle variations in language and linguistics, even when the primary language is English. ## The AfriMed-QA Dataset Composition * The dataset contains approximately 15,000 clinically diverse questions and answers sourced from 16 African countries. * It covers 32 medical specialties, ranging from neurosurgery and internal medicine to infectious diseases and obstetrics. * The content is divided into three distinct formats: 4,000+ expert multiple-choice questions (MCQs), 1,200 open-ended short-answer questions (SAQs), and 10,000 consumer-style queries. * Data was crowdsourced from 621 contributors across 60 medical schools to ensure a broad representation of the continent's medical landscape. ## Data Collection and Curation Methodology * Researchers adapted a specialized web-based platform, originally built by Intron Health, to facilitate large-scale crowdsourcing across different regions. * To protect privacy, consumer queries were generated by prompting users with specific disease scenarios rather than asking for personal health information. * The curation process included custom user interfaces for quality reviews and blinded human evaluations by clinical experts to ensure the accuracy of reference answers. ## LLM Performance and Evaluation Results * The study benchmarked 30 general and biomedical LLMs, evaluating them for accuracy, semantic similarity, and human preference. * A significant performance gap exists between model sizes; larger models consistently outperformed smaller models on the AfriMed-QA benchmark. * This trend highlights a challenge for low-resource settings, where smaller, specialized models are often preferred for on-device or edge deployment due to infrastructure constraints. * The dataset has already been utilized to improve Google’s MedGemma, demonstrating its utility in training multimodal medical models. The AfriMed-QA benchmark datasets and evaluation code have been open-sourced on Hugging Face and GitHub to support the global research community. Developers are encouraged to use these tools to build and refine medical AI that is more inclusive and effective for the Global South.

Natural Language Processing AI Large Language Models Benchmarking+5

lineAug 29, 2025Original article

Extracting Trending Keywords from OpenChat (opens in new tab)

To enhance user engagement on the LINE OpenChat main screen, LY Corporation developed a system to extract and surface "trending keywords" from real-time message data. By shifting focus from chat room recommendations to content-driven keyword clusters, the team addresses the lack of context in individual messages while providing a more dynamic discovery experience. This approach utilizes a combination of statistical Z-tests to identify frequency spikes and MinHash clustering to eliminate near-duplicate content, ensuring that the trending topics are both relevant and diverse. **The Shift from Chat Rooms to Content-Driven Recommendations** * Traditional recommendations focus on entire chat rooms, which often require significant user effort to investigate and evaluate. * Inspired by micro-blogging services, the team aimed to surface messages as individual content pieces to increase the "main screen visit" KPI. * Because individual chat messages are often fragmented or full of typos, the system groups them by keywords to create meaningful thematic content. **Statistical Detection of Trending Keywords** * Simple frequency counts are ineffective because they capture common social fillers like greetings or expressions of gratitude rather than actual trends. * Trends are defined as keywords showing a sharp increase in frequency compared to a baseline from seven days prior. * The system uses a Z-test for two-sample proportions to assign a score to each word, filtering for terms with at least a 30% frequency growth. * A seven-day comparison window is specifically used to suppress weekly cyclical noise (e.g., mentions of "weekend") and to capture topics whose popularity peaks over several consecutive days. **MinHash-based Message Deduplication** * Redundant messages, such as copy-pasted text, are removed prior to frequency aggregation to prevent skewed results and repetitive user experiences. * The system employs MinHash, a dimensionality reduction technique, to identify near-duplicate messages based on Jaccard similarity. * The process involves "shingling" messages into sets of tokens (primarily nouns) and generating $k$-length signatures; messages with identical signatures are clustered together. * To evaluate the efficiency of these clusters without high computational costs, the team developed a "SetDiv" (Set Diversity) metric that operates in linear time complexity. By combining Z-test statistical modeling with MinHash deduplication, this methodology successfully transforms fragmented chat data into a structured discovery layer. For developers working with high-volume social data, using a rolling weekly baseline and signature-based clustering offers a scalable way to surface high-velocity trends while filtering out both routine social noise and repetitive content.

Natural Language Processing AI Machine Learning Statistical Analysis+5

googleAug 25, 2025Original article

A scalable framework for evaluating health language models (opens in new tab)

Researchers at Google have developed a scalable framework for evaluating health-focused language models by replacing subjective, high-complexity rubrics with granular, binary criteria. This "Adaptive Precise Boolean" approach addresses the high costs and low inter-rater reliability typically associated with expert-led evaluation in specialized medical domains. By dynamically filtering rubric questions based on context, the framework significantly improves both the speed and precision of model assessments. ## Limitations of Traditional Evaluation * Current evaluation practices for health LLMs rely heavily on human experts, making them cost-prohibitive and difficult to scale. * Standard tools, such as Likert scales (e.g., 1-5 ratings) or open-ended text, often lead to subjective interpretations and low inter-rater consistency. * Evaluating complex, personalized health data requires a level of detail that traditional broad-scale rubrics fail to capture accurately. ## Precise Boolean Rubrics * The framework "granularizes" complex evaluation targets into a larger set of focused, binary (Yes/No) questions. * This format reduces ambiguity by forcing raters to make definitive judgments on specific aspects of a model's response. * By removing the middle ground found in multi-point scales, the framework produces a more robust and actionable signal for programmatic model refinement. ## The Adaptive Filtering Mechanism * To prevent the high volume of binary questions from overwhelming human raters, the researchers introduced an "Adaptive" layer. * The framework uses the Gemini model as a zero-shot classifier to analyze the user query and LLM response, identifying only the most relevant rubric questions. * This data-driven adaptation ensures that human experts only spend time on pertinent criteria, resulting in "Human-Adaptive Precise Boolean" rubrics. ## Performance and Reliability Gains * The methodology was validated in the domain of metabolic health, covering topics like diabetes, obesity, and cardiovascular disease. * The Adaptive Precise Boolean approach reduced human evaluation time by over 50% compared to traditional Likert-scale methods. * Inter-rater reliability, measured through intra-class correlation coefficients (ICC), was significantly higher than the baseline, proving that simpler scoring can provide a higher quality signal. This framework demonstrates that breaking down complex medical evaluations into simple, machine-filtered binary questions is a more efficient path toward safe and accurate health AI. Organizations developing domain-specific models should consider adopting adaptive binary rubrics to balance the need for expert oversight with the requirements of large-scale model iteration.

Natural Language Processing AI Large Language Models Gemini+4

googleAug 11, 2025Original article

Enabling physician-centered oversight for AMIE (opens in new tab)

Guardrailed-AMIE (g-AMIE) is a diagnostic AI framework designed to perform patient history-taking while strictly adhering to safety guardrails that prevent it from providing direct medical advice. By decoupling data collection from clinical decision-making, the system enables an asynchronous oversight model where primary care physicians (PCPs) review and finalize AI-generated medical summaries. In virtual clinical trials, g-AMIE’s diagnostic outputs and patient communications were preferred by overseeing physicians and patient actors over human-led control groups. ## Multi-Agent Architecture and Guardrails * The system utilizes a multi-agent setup powered by Gemini 2.0 Flash, consisting of a dialogue agent, a guardrail agent, and a SOAP note agent. * The dialogue agent conducts history-taking in three distinct phases: general information gathering, targeted validation of a differential diagnosis, and a conclusion phase for patient questions. * A dedicated guardrail agent monitors and rephrases responses in real-time to ensure the AI abstains from sharing individualized diagnoses or treatment plans directly with the patient. * The SOAP note agent employs sequential multi-step generation to separate summarization tasks (Subjective and Objective) from more complex inferential tasks (Assessment and Plan). ## The Clinician Cockpit and Asynchronous Oversight * To facilitate human review, researchers developed the "clinician cockpit," a web interface co-designed with outpatient physicians through semi-structured interviews. * The interface is structured around the standard SOAP note format, presenting the patient’s perspective, measurable data, differential diagnosis, and proposed management strategy. * This framework allows overseeing PCPs to review cases asynchronously, editing the AI’s proposed differential diagnoses and management plans before sharing a final message with the patient. * The separation of history-taking from decision-making ensures that licensed medical professionals retain ultimate accountability for patient care. ## Performance Evaluation via Virtual OSCE * The system was evaluated in a randomized, blinded virtual Objective Structured Clinical Examination (OSCE) involving 60 case scenarios. * g-AMIE’s performance was compared against primary care physicians, nurse practitioners, and physician assistants who were required to operate under the same restrictive guardrails. * Overseeing PCPs and independent physician raters preferred g-AMIE’s diagnostic accuracy and management plans over those of the human control groups. * Patient actors reported a preference for the messages generated by g-AMIE compared to those drafted by human clinicians in the study. While g-AMIE demonstrates high potential for human-AI collaboration in diagnostics, the researchers emphasize that results should be interpreted with caution. The workflow was specifically optimized for AI characteristics, and human clinicians may require specialized training to perform effectively within such highly regulated guardrail frameworks.

Natural Language Processing AI Generative AI Multi Agent Systems+4

googleJun 26, 2025Original article

REGEN: Empowering personalized recommendations with natural language (opens in new tab)

Google Research has introduced REGEN, a benchmark dataset designed to evolve recommender systems from simple item predictors into conversational agents capable of natural language interaction. By augmenting the Amazon Product Reviews dataset with synthetic critiques and narratives using Gemini 1.5 Flash, the researchers provide a framework for training models to understand user feedback and explain their suggestions. The study demonstrates that integrating natural language critiques significantly improves recommendation accuracy while enabling models to generate personalized, context-aware content. ### Composition of the REGEN Dataset * The dataset enriches the existing Amazon Product Reviews archive by adding synthetic conversational elements, specifically targeting the gap in datasets that support natural language feedback. * **Critiques** are generated for similar item pairs within hierarchical categories, allowing users to guide the system by requesting specific changes, such as a different color or increased storage. * **Narratives** provide contextual depth through purchase reasons, product endorsements, and concise user summaries, helping the system justify its recommendations to the end-user. ### Unified Generative Modeling Approaches * The researchers framed a "jointly generative" task where models must process a purchase history and optional critique to output both a recommended item ID and a supporting narrative. * The **FLARE (Hybrid)** architecture uses a sequential recommender for item prediction based on collaborative filtering, which then feeds into a Gemma 2B LLM to generate the final text narrative. * The **LUMEN (Unified)** model functions as an end-to-end system where item IDs and text tokens are integrated into a single vocabulary, allowing one LLM to handle critiques, recommendations, and narratives simultaneously. ### Performance and Impact of User Feedback * Incorporating natural language critiques consistently improved recommendation metrics across different architectures, demonstrating that language-guided refinement is a powerful tool for accuracy. * In the Office domain, the FLARE hybrid model's Recall@10—a measure of how often the desired item appears in the top 10 results—increased from 0.124 to 0.1402 when critiques were included. * Results indicate that models trained on REGEN can achieve performance comparable to state-of-the-art specialized recommenders while maintaining high-quality natural language generation. The REGEN dataset and the accompanying LUMEN architecture provide a path forward for building more transparent and interactive AI assistants. For developers and researchers, utilizing these conversational benchmarks is essential for moving beyond "black box" recommendations toward systems that can explain their logic and adapt to specific user preferences in real time.

Natural Language Processing AI Large Language Models Gemini+4

googleMay 5, 2025Original article

Making complex text understandable: Minimally-lossy text simplification with Gemini (opens in new tab)

Google Research has introduced a novel system using Gemini models to perform minimally-lossy text simplification, a process designed to enhance readability while meticulously preserving original meaning and nuance. By utilizing an automated, iterative prompt-refinement loop, the system optimizes LLM instructions to achieve high-fidelity paraphrasing that avoids the information loss typical of standard summarization. A large-scale randomized study confirms that this approach significantly improves user comprehension across complex domains like law and medicine while simultaneously reducing cognitive load for the reader. ## Automated Evaluation and Fidelity Assessment * The system moves beyond traditional metrics like Flesch-Kincaid by using a Gemini-powered 1-10 readability scale that aligns more closely with human judgment and comprehension ease. * Fidelity is maintained through a specialized process using Gemini 1.5 Pro that maps specific claims from the original source text directly to the simplified output. * This mapping method identifies and weights specific error types, such as information loss, unnecessary gains, or factual distortions, to ensure the output remains a faithful representation of the technical original. ## Iterative Prompt Optimization Loop * To overcome the limitations and speed of manual prompt engineering, the researchers implemented a feedback loop where Gemini models optimize their own instructions. * In this "LLMs optimizing LLMs" setup, Gemini 1.5 Pro analyzes the performance of simplification prompts and proposes refinements based on automated readability and fidelity scores. * The optimization process ran for 824 iterations before performance plateaued, allowing the system to autonomously discover highly effective strategies for simplifying text without sacrificing detail. ## Validating Impact through Randomized Studies * The effectiveness of the model was validated with 4,563 participants across 31 diverse text excerpts covering specialized fields like aerospace, philosophy, finance, and biology. * The study utilized a randomized complete block design to compare the original text against simplified versions, measuring outcomes through nearly 50,000 multiple-choice question responses. * Beyond accuracy, researchers measured cognitive effort using the NASA Task Load Index and tracked self-reported user confidence to ensure the simplification actually lowered the barrier to understanding. This technology provides a scalable method for democratizing access to specialist knowledge by making expert-level discourse understandable to a general audience. The system is currently available as the "Simplify" feature within the Google app for iOS, offering a practical tool for users navigating complex digital information.

Natural Language Processing AI Large Language Models Gemini+3

googleMay 1, 2025Original article

Amplify Initiative: Localized data for globalized AI (opens in new tab)

The Amplify Initiative by Google Research addresses the critical lack of linguistic and cultural diversity in generative AI training data by establishing an open, community-based platform for localized data collection. By partnering with regional experts to co-create structured, high-quality datasets, the initiative aims to ensure AI models are both representative and effective in solving local challenges across health, finance, and education. This approach shifts data collection from a top-down model to a participatory framework that prioritizes responsible, locally respectful practices in the Global South. ## The Amplify Platform Framework The initiative is designed to bridge the gap between global AI capabilities and local needs through three core pillars: * **Participatory Co-creation:** Researchers and local communities collaborate to define specific data needs, ensuring the resulting datasets address region-specific problems like financial literacy or localized health misinformation. * **Open Access for Innovation:** The platform provides high-quality, multilingual datasets suitable for fine-tuning and evaluating models, specifically empowering developers in the Global South to build tools for their own communities. * **Author Recognition:** Contributors receive tangible rewards, including professional certificates, research acknowledgments, and data authorship attribution, creating a sustainable ecosystem for expert participation. ## Pilot Implementation in Sub-Saharan Africa To test the methodology, Google Research partnered with Makerere University’s AI Lab in Uganda to conduct an on-the-ground pilot program. * **Expert Onboarding:** The program trained 259 experts across Ghana, Kenya, Malawi, Nigeria, and Uganda through a combination of in-person workshops and app-based modules. * **Dataset Composition:** The pilot resulted in 8,091 annotated adversarial queries across seven languages, covering salient domains such as education and finance. * **Adversarial Focus:** By focusing on adversarial queries, the team captured localized nuances of potential AI harms, including regional stereotypes and specialized advice that generic models often miss. ## Technical Workflow and App-Based Methodology The initiative utilizes a structured technical pipeline to scale data collection while maintaining high quality and privacy. * **Privacy-Preserving Android App:** A dedicated app serves as the primary interface for training, data creation, and annotation, allowing experts to contribute from their own environments. * **Automated Validation:** The app includes built-in feedback loops that use automated checks to ensure queries are relevant and to prevent the submission of semantically similar or duplicate entries. * **Domain-Specific Annotation:** Experts are provided with specialized annotation topics tailored to their professional backgrounds, ensuring that the metadata for each query is technically accurate and contextually relevant. The Amplify Initiative provides a scalable blueprint for building inclusive AI by empowering experts in the Global South to define their own data needs. As the project expands to India and Brazil, it offers a vital resource for developers seeking to fine-tune models for local contexts and improve the safety and relevance of AI on a global scale.

Natural Language Processing AI Large Language Models Generative AI+5