Activation Steering: One Dial, Two Costs
Activation steering edits a model's internal activations to add or remove a concept at inference time. This explainer maps its one control knob, its two costs, and when it beats prompting.
Loading…
Topic
LLMs
648 posts about LLMs. Every summary links to the original.
- 648 posts
- 49 companies
- 5 explainers
See the idea
LLMs explainers
Antibodies as a Prediction Pipeline: Where AI Models Break Down
Antibody discovery is a three-stage prediction pipeline, and AI models that excel on familiar targets degrade sharply on unfamiliar ones. This explainer maps where that gap appears and what it costs.
Agent Memory: What Outlives the Run
Agent memory is the knowledge that survives a run, not the working state inside one. This explainer separates the two, then maps the write path, the read path, and the database limits that cap recall.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
Enterprise-Managed Authorization for your MCP server: Setting Up and Testing XAA in Auth0
Connecting AI agents to enterprise APIs and MCP servers creates governance challenges when relying on traditional OAuth consent prompts. Cross App Access shifts authorization decisions to the enterprise identity provider, allowing IT administrators to manage agent permissions centrally. In this architecture, an identity provider like Okta authenticates the user and issues an Identity Assertion JWT Authorization Grant based on administrative policy. The requesting AI agent presents this assertion to Auth0, which validates the signature and exchanges it for a standard access token via a JWT-bearer grant. Protected APIs can then validate the final token normally without requiring code changes or direct user consent interactions.
Multilingual Knowledge Transfer under Data Constraints via Lexical Interventions
Cross-lingual knowledge transfer is critical for training multilingual language models when target-language data is scarce and downstream tasks demand reasoning or world knowledge. Existing transfer approaches often depend on scarce resources such as large parallel corpora, translation systems, auxiliary models, or extra training phases. To address this limitation, the authors introduce LINK, a data-level intervention method that operates directly during pretraining via bilingual vocabularies. LINK swaps randomly selected words in a portion of the high-resource English corpus with word-level translations according to a specified replacement ratio. Across evaluations spanning eight languages and five model sizes, the technique delivered notable performance gains on downstream target-language tasks while achieving up to a 2x speedup in training to reach equivalent performance.
Scaling Laws for Mixture Pretraining Under Data Constraints
Scaling language models demands large volumes of data, but high-value target sources like low-resource languages and specialized domains remain inherently scarce. Practitioners frequently combine limited target data with abundant generic data, creating a balance between domain underexposure and excessive repetition leading to overfitting. An empirical evaluation across more than 2,000 language-model training runs demonstrates that mixture training tolerates substantially higher data repetition than single-source training, allowing scarce target corpora to be repeated 15 to 20 times. Researchers introduced a repetition-aware mixture scaling law that accounts for both the diminishing returns of repeated target tokens and the regularizing effect of generic data. Optimizing this scaling law offers a principled framework for calculating effective mixture configurations and practical pretraining recommendations under severe data constraints.
Developing NVIDIA Holoscan Applications with CLI, Skills, and AI Coding Agents
NVIDIA Holoscan is a platform for real-time AI applications at the edge, and HoloHub provides reference applications and components. This post describes an engineer-guided workflow in which a coding agent builds a separate Python endoscopic tool segmentation application from the existing MONAI model, sample video, and Holoscan patterns. The Holoscan CLI, invoked through ./holohub, exposes shared commands for scaffolding, running, modes, tests, and benchmarking, while development skills and documentation guide the agent through project-specific operations. The application connects video replay, preprocessing, TensorRT inference, segmentation postprocessing, telemetry, and HoloViz visualization, with visual, smoke, and benchmark modes; the post reports that the CLI-plus-skills-plus-docs/examples configuration produced a separate application in 40 minutes using 11M tokens, and concludes that this combination offered the lowest resource overhead among the compared configurations.
How we think about text classification in the LLM era
Medium revisited its aging NSFW text-classification model, which had outdated code, infrequent retraining, and poorly documented training and performance history. The replacement uses an LLM prompt containing NSFW guidelines and story content, requesting a 0-to-100 score, structured output, and a short reason rather than requiring a training set. To evaluate it, the team created a curator-labeled dataset with deliberate oversampling of NSFW stories, then compared several LLMs with the legacy model using false-positive and false-negative metrics. Some models matched or exceeded the legacy model; the team selected Mixtral 8×7b based on cost and performance and ran inference in Snowflake. The post presents this as a strong fit where training data is costly, trends are relatively stable, and only distribution-eligible stories need scoring, while topic and spam models still benefit from continuously updated datasets.
Designing effective Genie Agents from a single prompt
Generic agents may select the first relevant-looking revenue table rather than the finance-maintained source because they lack business context. Genie Agents address this by using a single prompt to assemble a domain-specific agent from governed Unity Catalog context, including structured data, documents, and files. The prompt identifies the desired outcome and relevant sources, while the agent can retrieve content from PDFs, Word documents, presentations, images, and tables under the requesting user’s permissions. The recommended process is to begin with one focused, repetitive job, test it against known questions or incidents, and expand only after measuring performance. Built-in benchmarks provide accuracy scores, while monitoring surfaces user questions and feedback that can reveal gaps in the curated context.
Evaluating AI Agent Skill Performance with NVIDIA SkillEvaluator
NVIDIA SkillEvaluator measures whether packaged agent skills improve agent performance by comparing task runs with and without a skill installed. It combines three evaluation tiers: static safety and structure checks, embedding-based distinctiveness analysis, and live evaluations in isolated Harbor sandboxes. In the August 12, 2026 benchmark snapshot, more than 300 verified skills spanning over 30 NVIDIA products were tested across Claude Code and Codex, with macro-averaged Skill Lift of +31 points across all dimensions and +39 excluding Security. Correctness rose from 46 to 87 and Effectiveness from 39 to 78, while Security moved from 97 to 98; token and runtime effects varied by skill, with one reducing both and another increasing them.
How Clario technology detects PHI/PII in DICOM images using Amazon Bedrock
Clario developed an automated PHI and PII detection solution on AWS to process thousands of DICOM image slices and PDF files in clinical trials. Medical imaging workflows require strict privacy compliance under HIPAA, GDPR, and ICH E6 guidelines across standard tags, custom vendor metadata, and pixel-burned text. To address this, the pipeline retrieves files from Amazon S3 and runs memory-intensive detection workloads on Amazon EKS backed by Amazon RDS for PostgreSQL. Amazon Textract handles optical character recognition, while Anthropic's Claude Sonnet on Amazon Bedrock classifies sensitive data using domain-optimized prompt engineering. The system identifies precise coordinates and sensitive entity types rather than directly altering files, enabling human-in-the-loop review before downstream redaction occurs.
AI-powered clinical trial eligibility and safety using Amazon Bedrock AgentCore
Manual chart review across fragmented clinical sources slows clinical trial enrollment, contributing to timeline delays and high screen failure rates. To address this bottleneck, an architecture on AWS automates eligibility and safety assessments while preserving human clinician authority. AWS HealthLake ingests and normalizes records into FHIR R4 resources, while Amazon Bedrock AgentCore orchestrates pre-screening, detailed screening, and site enrollment agents. Amazon Bedrock AgentCore Evaluations scores screening determinations using an LLM-as-a-judge approach for clinical accuracy, operational effectiveness, and safety compliance. This human-in-the-loop workflow produces structured recommendations with source citations, reducing patient matching time from days to minutes while maintaining compliance audit trails.
Accelerate automation with AI and the Ansible development tools MCP servers
Ansible automation development often involves frequent context switching and struggles with AI assistants generating syntactically valid code that hallucinates module parameters or misses local dependencies. To address this gap, the Ansible development tools Model Context Protocol server connects AI assistants to external development tooling through structured interfaces. The server wraps command-line utilities, enabling AI models to deterministically run ansible-lint with automatic fixes, scaffold projects via ansible-creator, manage environments, and build execution environment container images. By surfacing local workspace schemas, documentation, and organizational linting standards, the system grounds AI responses in real environment configurations. The integration operates across multiple MCP-compatible clients without requiring a specific model provider.
Gisting: Compressing LLM Agent context to ↑ throughput and ↓ cost
Long system prompts improve agent behavior but increase inference time, cost, and the GPU capacity needed to serve a given traffic level. Shopify compresses the Sidekick GraphQL agent’s roughly 6,000-token system prompt into about 1,500 gist tokens, using special vocabulary tokens whose embeddings are learned by knowledge distillation while the model weights remain frozen. Training compares teacher logits from the full prompt with student logits from the gist-token prompt using KL divergence; at inference, the prompt is replaced with the gist-token string, requiring no custom serving path. In load tests at 350 requests per minute, compression reduced median time to first token from 438ms to 354ms and end-to-end latency from 6.8s to 4.2s, while raising throughput from 20.2 to 23.4 QPS and enabling 14% fewer GPUs in production without losing prediction quality.
Building Secure AI Agents with Microsoft Agent Framework and Auth0: Sending Email with Token Vault
An expense-reviewing AI agent required manual manager intervention to email employees about missing report data such as amounts, merchants, or justifications. To automate follow-ups safely, the agent uses Auth0 Token Vault to acquire and refresh Google OAuth tokens without storing credentials in the application database. The implementation configures Auth0 with offline access, My Account API permissions, and multi-resource refresh tokens to handle silent token retrieval across user sessions. Rather than exposing Gmail credentials to the underlying language model, the agent executes email actions via an internal backend endpoint that retrieves a short-lived access token, sends the email, and immediately discards the token.
Examining Human-Like Behaviors in LLMs: A Multi-Dimensional Analysis of Model Behaviors, User Factors, and System Prompts
Large language models frequently exhibit human-like behaviors such as expressing emotions, building relationships, and setting boundaries, yet empirical guidance on appropriate usage remains limited. To address this gap, researchers analyzed 21,000 multi-turn conversations across four models: gpt-4o, gpt-4.1-mini, claude-sonnet-4.6, and gemini-2.5-flash. The study combined LLM-as-a-judge techniques with human evaluation to measure behavior prevalence, perceived appropriateness, and prompt controllability across diverse user goals and profiles. Human evaluators judged LLM self-referential and relationship-building behaviors as less appropriate than identical human actions, whereas boundary-maintaining behaviors were rated as more appropriate for models. System prompting effectively steered these behaviors, though the authors emphasize that rigorous evaluation remains necessary to prevent unintended side effects.
Databricks Document Intelligence: pushing the frontier for complex document extraction
Databricks introduces Precision Mode for its ai_extract document extraction API, targeting enterprise workloads that challenge existing LLM and rules-based systems. The mode addresses long documents with cross-references, large nested outputs such as thousands of line items, and schemas requiring reasoning across financial statements or applying discounts. It combines custom-trained, efficient extraction models with an agent harness that decomposes jobs semantically, runs smaller tasks in parallel, preserves intermediate results, and reconciles them into a final structured output. On benchmarks covering roughly 9,000 documents, including files up to 2,000 pages and schemas with more than 300 deeply nested fields, Precision Mode reached 94.7% accuracy, seven points above the strongest frontier-model chunk-and-merge baseline, GPT-5.6 Sol; the source says it is now available through ai_extract and the Information Extraction UI.
How Much Memory Does Your Agent Actually Need?
The post examines how much agentic memory different models can use effectively, evaluating ALTK-Evolve's self-distilled guidelines across eight models on AppWorld's 585 multi-step tasks. Memory is produced by extracting and consolidating lessons from successful and unsuccessful trajectories, then delivering either the full guideline set on every ReAct step or a fixed core plus task-relevant retrieval, without updating model weights. Results identify three patterns: strong models with headroom benefited from the full set, gpt-oss-120b gained 16.1 percentage points in task completion from curated retrieval, and GLM-5 showed no measurable gain; stricter scenario completion sometimes improved more. Curated retrieval gave gpt-oss-120b that gain with only 5% more tokens versus 51% for its full set, while prompt caching is presented as a way to reduce recurring input cost, supporting calibration to model capability rather than simply accumulating memory.
17,600 Actions: Agent Security Is a Systems Problem
Security incidents involving AI agents, such as the OpenAI and Hugging Face event where attackers executed 17,600 actions, demonstrate that system security cannot rely entirely on manual human review. Managing agent velocity requires dedicated system-level controls to properly constrain, observe, and govern their behaviors during execution. To secure autonomous development workflows, Docker Sandboxes allow AI agents to operate within GitHub Actions, run Testcontainers tests, adjust code, and automatically generate draft pull requests inside isolated environments. In addition, escalating software supply-chain threats and the increased volume of AI-generated code drive the adoption of Docker Hardened Images and Docker Extended Lifecycle Support. These updates incorporate software built from source into container images, extend security patch coverage past end-of-life milestones, and move organizational policy enforcement directly onto developer machines.
When it comes to Governance, Retailers need a control plane for context
Retailers are moving from isolated generative-AI pilots to enterprise-wide adoption, where each store, headquarters, and digital workflow needs different data, permissions, models, and cost controls. The post argues that useful AI depends on governed business context, while fragmented tools can produce separate provider contracts, logs, permission models, cost centers, and answers to the same question. It proposes an open control plane that centralizes governance and model access without forcing employees into one application or vendor. The described Databricks design combines Unity Catalog for data and business context, Unity AI Gateway for routing and oversight, Foundation Model APIs for commercial, open-source, and custom models, and Genie for natural-language queries over governed data; the intended result is faster experimentation and deployment with visibility into access, spending, lineage, and auditability.
Evaluating AI Agents Live at the Grounded Reasoning Cup
The Grounded Reasoning Cup evaluated whether AI agents could generalize grounded reasoning improvements from OfficeQA to a newly released, enterprise-style document benchmark. Eleven academic teams had about two months to build agents with a partner lab’s model family, then applied them live to OfficeQA Pro V2, based on U.S. Treasury Accounts of Receipts and Expenditures, after a 36-hour release window. Stanford won with 63.3% accuracy, while the top systems combined document preprocessing, targeted or hybrid retrieval, specialized tools, parallelism, and answer verification; 18.8% of questions remained unsolved by every team. The results show that benchmark gains did not reliably transfer, and that parsing, retrieval, tool use, verification, and operational infrastructure can matter as much as model choice in end-to-end enterprise reasoning.
Coding Agent Horror Stories: The Command You Already Approved
The OpenAI and Hugging Face security incident involved 17,600 attacker actions, demonstrating that AI agent security cannot rely exclusively on manual human review. Modern development workflows require systems controls to constrain, observe, and govern autonomous agents operating at high speeds. To address related isolation and automation requirements, developers can run AI agents inside GitHub Actions using Docker Sandboxes to execute Testcontainers tests, fix code, and generate draft pull requests. In addition, Docker Extended Lifecycle Support provides continued security coverage past software end of life while moving policy enforcement directly to developer workstations. Finally, Docker Hub has transitioned Docker Verified Publisher applications to a self-serve model, enabling publishers to distribute verified content directly to developers.
Consistency is the new latency: AI at the data layer
Autonomous AI agents depend on the accuracy of the underlying data layer because retrieved data populates the LLM context window for multi-step reasoning. In distributed architectures, standard asynchronous replication lag causes agents to ingest stale data, leading to flawed execution and self-reinforcing hallucination debt when errors are saved back to memory. To resolve this, architectures must align replication strategies with specific task truth requirements across three primary patterns. High-stakes workloads utilize Amazon Aurora with global write forwarding or native Aurora DSQL for multi-region strong consistency, while high-scale scenarios leverage DynamoDB Global Tables with conditional writes to prevent lost updates. Finally, high-velocity ingestion pipelines enforce LOCAL_QUORUM reads and writes in Amazon Keyspaces to capture critical data spikes reliably.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.