Tracing model outputs to the training data | Anthropic | Research | TechiSeek

AddAdd your Listings

Research

Tracing model outputs to the training data

Scaling influence functions to 52B-parameter models shows generalization becomes more abstract and cross-lingual as models grow.

Research Anthropic

ResearchDiscovering behaviors with model-written evaluationsAnthropic ResearchTraining a helpful and harmless assistant with RLHFAnthropic ResearchSleeper Agents: Training Deceptive LLMs that Persist Through Safety TrainingAnthropic ResearchOpen-sourcing circuit-tracing toolsAnthropic ResearchGLM-5.3 and the spread of advanced cyber capabilitiesAnthropic ResearchMeasuring LLMs’ ability to develop exploitsAnthropic

EventsCloud Security Trends & Challenges: Complete GuideCybersecurity Exchange BlogsAI Adoption for National Security | CohereCohere BlogsIntroducing ClickStack Cloud: Serverless observability powered by ClickHouseClickhouse Products & Services1Password Insights - Use Watchtower Reports to Identify Risks1password Products & ServicesDevice Security Ensures Secure Access in Hybrid Workforce1password Products & ServicesExtended Device Compliance: Secure Apps Beyond SSO1password LearnFree "DuckDB in Action" BookMotherduck EventsTime Travel & Schema Evolution: DuckLake Handles ChangeMotherduck EventsJev for Analytics: 5 Use Cases on Real DataMotherduck BlogsWhy CSV Files Won’t Die and How DuckDB Conquers ThemMotherduck BlogsThis Month in the DuckDB Ecosystem: July 2025Motherduck EventsPyData Amsterdam Meetup: Agents in the Data-verseMotherduck BlogsFuture Casting the Modern Data StackMotherduck EventsBoosting Data Performance: Unlocking the Power of DuckDB in your Gold LayerMotherduck