Events & Conferences
Updating large language models by directly editing network layers
One of the great attractions of large language models (LLMs) is that they encode information about the real world. But the world is constantly changing, and an LLM’s information is only as fresh as the data it was trained on.
Training an LLM can take months, even when the task is parallelized across 1,000 servers, so AI researchers have sought alternate ways to update LLMs’ knowledge. One of these is to directly edit targeted layers of an LLM, to improve its performance on a particular knowledge-based task. This is a task-specific solution, not a general solution, but it takes hours to implement rather than months.
Existing techniques for direct layer editing generally require either manual selection of layers to be edited or a time-consuming procedure to determine the layers where editing will do the most good. Last week, at the 2024 meeting of the European Chapter of the Association for Computational Linguistics, we presented a new method for automatically selecting layers to be edited, which yields more-accurate updates than previous automated methods.
Compared to the prior method for manual layer selection, it also limits regression, or post-update backsliding on data that the model previously handled correctly. On some datasets, our method, which we call SaLEM (for salient-layers editing model), reduced regression by an order of magnitude, while offering equivalent accuracy on new data.
Identifying layers
We consider the case in which an LLM has been fine-tuned on a specific task, such as determining whether one input sentence logically entails or counts as evidence for or against another. In such cases, the model input is typically a pair of texts, and the output is a decision such as “entailed” or “supported”.
In the prior approach to manual layer selection, known as causal tracing, the first token of each training example is fed to the model, then the first and second, then the first, second, and third, and so on. Then the process is repeated with one of the model layers masked. This two-step analysis, in turn, must be repeated for each layer of the network, a time-consuming procedure.
In our case, we instead prepare an “edit dataset”, consisting of input-output pairs drawn from three groups: (1) the pass samples, for which the existing model outputs the correct answers; (2) the fail samples, for which the existing model outputs the wrong answers; and (3) the adapt samples, which are semantically equivalent to the fail samples but differently phrased.
For each sample, we compute the loss between the existing model’s output and the target output and the corresponding gradients — the modifications of model weights that make correct outputs more likely. Then we average the gradients across each layer of the model and across all training samples. The layer with highest average gradient — the layer that requires the largest modification to accommodate new facts about the world — is the one we edit.
Layer editing
To edit the selected layer, we use the MEND method proposed by Stanford University researchers in 2022. With MEND, a second machine learning model, the editor model, is trained to, essentially, take gradients as inputs and output parameter edits.
But rather than the raw gradients, the model’s inputs are a low-rank approximation of the gradients, which reduces the dimension of the data by identifying the axes along which most of the variance occurs. This is something like teasing out the underlying causes of the larger gradients, which helps the model generalize better. We also guard against overfitting by aggregating gradients in batches of 10 before computing their low-rank approximation.
We use two training objectives to train the editor, one that maximizes the likelihood of correct answers on inputs from the fail and adapt sets and one that minimizes output divergence on inputs from the pass set. This helps prevent regression.
In the original MEND paper, the Stanford researchers used this approach to edit the top three layers of a fine-tuned LLM, a reasonable heuristic for trading off editing efficiency, correction of outputs, and prevention of regression. Because SaLEM identifies the one layer most implicated in the new model update, it can match MEND’s performance on new data. But because it modifies parameters in one layer rather than three, it reduces regression.
Experiments
We evaluated SaLEM on six datasets used to fine-tune LLMs on natural-language-processing tasks. Four of the datasets had to do with natural-language inference, one was a question-answering dataset, and one was a dataset for the standard LLM task of next-token prediction. For the question-answering and generation tasks, we compared SaLEM and the baselines on four different LLM architectures. We measured performance using both edit accuracy, or post-editing accuracy on the new data, and drawdown, which measures regression on the old data.
On the inference tasks, SaLEM matched the edit accuracy of the top performers but had significantly better drawdown — four and ten times better than the second-best performer on two of the datasets. On the other two tasks, SaLEM finished second on both measures to an approach called editable neural networks (ENN). But ENN requires two copies of an LLM to run simultaneously, which is resource intensive. Indeed, for two of the four LLM architectures we tested, we were unable to run ENN because of its computational demands.
In ongoing work, we are investigating (1) enriching the editing dataset with better failed samples and their semantic and counterfactual equivalents, (2) a better weight update mechanism to inform the editor about the extent of updates for borderline instances, and (3) a method of performing edits without loading the full model into memory, as we currently do.
Have you ever worked is legacy code? Are you curious what it takes to modernize systems at a massive scale?
Pascal Hartig is joined on the latest Meta Tech Podcast by Elaine and Buping, two software engineers working on a bold project to rewrite the decades-old C code in one of Meta’s core messaging libraries in Rust. It’s an ambitious effort that will transform a central messaging library that is shared across Messenger, Facebook, Instagram, and Meta’s AR/VR platforms.
They discuss taking on a project of this scope – even without a background in Rust, how they’re approaching it, and what it means to optimize for ‘developer happiness.’
Download or listen to the episode below:
You can also find the episode wherever you get your podcasts, including:
The Meta Tech Podcast is a podcast, brought to you by Meta, where we highlight the work Meta’s engineers are doing at every level – from low-level frameworks to end-user features.
Send us feedback on Instagram, Threads, or X.
And if you’re interested in learning more about career opportunities at Meta visit the Meta Careers page.
Amazon Research Awards (ARA) provides unrestricted funds and AWS Promotional Credits to academic researchers investigating various research topics in multiple disciplines. This cycle, ARA received many excellent research proposals from across the world and today is publicly announcing 73 award recipients who represent 46 universities in 10 countries.
This announcement includes awards funded under five call for proposals during the fall 2024 cycle: AI for Information Security, Automated Reasoning, AWS AI, AWS Cryptography, and Sustainability. Proposals were reviewed for the quality of their scientific content and their potential to impact both the research community and society. Additionally, Amazon encourages the publication of research results, presentations of research at Amazon offices worldwide, and the release of related code under open-source licenses.
Recipients have access to more than 700 Amazon public datasets and can utilize AWS AI/ML services and tools through their AWS Promotional Credits. Recipients also are assigned an Amazon research contact who offers consultation and advice, along with opportunities to participate in Amazon events and training sessions.
“Automated Reasoning is an important area of research for Amazon, with potential applications across various features and applications to help improve security, reliability, and performance for our customers. Through the ARA program, we collaborate with leading academic researchers to explore challenges in this field,” said Robert Jones, senior principal scientist with the Cloud Automated Reasoning Group. “We were again impressed by the exceptional response to our Automated Reasoning call for proposals this year, receiving numerous high-quality submissions. Congratulations to the recipients! We’re excited to support their work and partner with them as they develop new science and technology in this important area.”
“At Amazon, we believe that solving the world’s toughest sustainability challenges benefits from both breakthrough scientific research and open and bold collaboration. Through programs like the Amazon Research Awards program, we aim to support academic research that could contribute to our understanding of these complex issues,” said Kommy Weldemariam, Director of Science and Innovation Sustainability. “The selected proposals represent innovative projects that we hope will help advance knowledge in this field, potentially benefiting customers, communities, and the environment.”
ARA funds proposals throughout the year in a variety of research areas. Applicants are encouraged to visit the ARA call for proposals page for more information or send an email to be notified of future open calls.
The tables below list, in alphabetical order by last name, fall 2024 cycle call-for-proposal recipients, sorted by research area.
AI for Information Security
| Recipient | University | Research title |
| Christopher Amato | Northeastern University | Multi-Agent Reinforcement Learning Cyber Defense for Securing Cloud Computing Platforms |
| Bernd Bischl | Ludwig Maximilian University of Munich | Improving Generative and Foundation Models Reliability via Uncertainty-awareness |
| Shiqing Ma | University Of Massachusetts Amherst | LLM and Domain Adaptation for Attack Detection |
| Alina Oprea | Northeastern University | Multi-Agent Reinforcement Learning Cyber Defense for Securing Cloud Computing Platforms |
| Roberto Perdisci | University of Georgia | ContextADBench: A Comprehensive Benchmark Suite for Contextual Anomaly Detection |
Automated Reasoning
| Recipient | University | Research title |
| Nada Amin | Harvard University | LLM-Augmented Semi-Automated Proofs for Interactive Verification |
| Suguman Bansal | Georgia Institute of Technology | Certified Inductive Generalization in Reinforcement Learning |
| Ioana Boureanu | University of Surrey | Phoebe+: An Automated-Reasoning Tool for Provable Privacy in Cryptographic Systems |
| Omar Haider Chowdhury | Stony Brook University | Restricter: An Automatic Tool for Authoring Amazon Cedar Access Control Policies with the Principle of Least Privilege |
| Stefan Ciobaca | Alexandru Ioan Cuza University | An Interactive Proof Mode for Dafny |
| João Ferreira | INESC-ID | Polyglot Automated Program Repair for Infrastructure as Code |
| Sicun Gao | University Of California, San Diego | Monte Carlo Trees with Conflict Models for Proof Search |
| Mirco Giacobbe | University of Birmingham | Neural Software Verification |
| Tobias Grosser | University of Cambridge | Synthesis-based Symbolic BitVector Simplification for Lean |
| Ronghui Gu | Columbia University | Scaling Formal Verification of Security Properties for Unmodified System Software |
| Alexey Ignatiev | Monash University | Huub: Next-Gen Lazy Clause Generation |
| Kenneth McMillan | University of Texas At Austin | Synthesis of Auxiliary Variables and Invariants for Distributed Protocol Verification |
| Alexandra Mendes | University of Porto | Overcoming Barriers to the Adoption of Verification-Aware Languages |
| Jason Nieh | Columbia University | Scaling Formal Verification of Security Properties for Unmodified System Software |
| Rohan Padhye | Carnegie Mellon University | Automated Synthesis and Evaluation of Property-Based Tests |
| Nadia Polikarpova | University Of California, San Diego | Discovering and Proving Critical System Properties with LLMs |
| Fortunat Rajaona | University of Surrey | Phoebe+: An Automated-Reasoning Tool for Provable Privacy in Cryptographic Systems |
| Subhajit Roy | Indian Institute of Technology Kanpur | Theorem Proving Modulo LLM |
| Gagandeep Singh | University of Illinois At Urbana–Champaign | Trustworthy LLM Systems using Formal Contracts |
| Scott Stoller | Stony Brook University | Restricter: An Automatic Tool for Authoring Amazon Cedar Access Control Policies with the Principle of Least Privilege |
| Peter Stuckey | Monash University | Huub: Next-Gen Lazy Clause Generation |
| Yulei Sui | University of New South Wales | Path-Sensitive Typestate Analysis through Sparse Abstract Execution |
| Nikos Vasilakis | Brown University | Semantics-Driven Static Analysis for the Unix/Linux Shell |
| Ping Wang | Stevens Institute of Technology | Leveraging Large Language Models for Reasoning Augmented Searching on Domain-specific NoSQL Database |
| John Wawrzynek | University of California, Berkeley | GPU-Accelerated High-Throughput SAT Sampling |
AWS AI
| Recipient | University | Research title |
| Panagiotis Adamopoulos | Emory University | Generative AI solutions for The Spillover Effect of Fraudulent Reviews on Product Recommendations |
| Vikram Adve | University of Illinois at Urbana–Champaign | Fellini: Differentiable ML Compiler for Full-Graph Optimization for LLM Models |
| Frances Arnold | California Institute of Technology | Closed-loop Generative Machine Learning for De Novo Enzyme Discovery and Optimization |
| Yonatan Bisk | Carnegie Mellon University | Useful, Safe, and Robust Multiturn Interactions with LLMs |
| Shiyu Chang | University of California, Santa Barbara | Cut the Crap: Advancing the Efficient Communication of Multi-Agent Systems via Spatial-Temporal Topology Design and KV Cache Sharing |
| Yuxin Chen | University of Pennsylvania | Provable Acceleration of Diffusion Models for Modern Generative AI |
| Tianlong Chen | University of North Carolina at Chapel Hill | Cut the Crap: Advancing the Efficient Communication of Multi-Agent Systems via Spatial-Temporal Topology Design and KV Cache Sharing |
| Mingyu Ding | University of North Carolina at Chapel Hill | Aligning Long Videos and Language as Long-Horizon World Models |
| Nikhil Garg | Cornell University | Market Design for Responsible Multi-agent LLMs |
| Jessica Hullman | Northwestern University | Human-Aligned Uncertainty Quantification in High Dimensions |
| Christopher Jermaine | Rice University | Fast, Trusted AI Using the EINSUMMABLE Compiler |
| Yunzhu Li | Columbia University | Physics-Informed Foundation Models Through Embodied Interactions |
| Pattie Maes | Massachusetts Institute of Technology | Understanding How LLM Agents Deviate from Human Choices |
| Sasa Misailovic | University of Illinois at Urbana–Champaign | Fellini: Differentiable ML Compiler for Full-Graph Optimization for LLM Models |
| Kristina Monakhova | Cornell University | Trustworthy extreme imaging for science using interpretable uncertainty quantification |
| Todd Mowry | Carnegie Mellon University | Efficient LLM Serving on Trainium via Kernel Generation |
| Min-hwan Oh | Seoul National University | Mutually Beneficial Interplay Between Selection Fairness and Context Diversity in Contextual Bandits |
| Patrick Rebeschini | University of Oxford | Optimal Regularization for LLM Alignment |
| Jose Renau | University of California, Santa Cruz | Verification Constrained Hardware Optimization using Intelligent Design Agentic Programming |
| Vilma Todri | Emory University | Generative AI solutions for The Spillover Effect of Fraudulent Reviews on Product Recommendations |
| Aravindan Vijayaraghavan | Northwestern University | Human-Aligned Uncertainty Quantification in High Dimensions |
| Wei Yang | University of Texas at Dallas | Optimizing RISC-V Compilers with RISC-LLM and Syntax Parsing |
| Huaxiu Yao | University of North Carolina at Chapel Hill | Aligning Long Videos and Language as Long-Horizon World Models |
| Amy Zhang | University of Washington | Tools for Governing AI Agent Autonomy |
| Ruqi Zhang | Purdue University | Efficient Test-time Alignment for Large Language Models and Large Multimodal Models |
| Zheng Zhang | Rutgers University-New Brunswick | AlphaQC: An AI-powered Quantum Circuit Optimizer and Denoiser |
AWS Cryptography
| Recipient | University | Research title |
| Alexandra Boldyreva | Georgia Institute of Technology | Quantifying Information Leakage in Searchable Encryption Protocols |
| Maria Eichlseder | Graz University of Technology, Austria | SALAD – Systematic Analysis of Lightweight Ascon-based Designs |
| Venkatesan Guruswami | University of California, Berkeley | Obfuscation, Proof Systems, and Secure Computation: A Research Program on Cryptography at the Simons Institute for the Theory of Computing |
| Joseph Jaeger | Georgia Institute of Technology | Analyzing Chat Encryption for Group Messaging |
| Aayush Jain | Carnegie Mellon | Large Scale Multiparty Silent Preprocessing for MPC from LPN |
| Huijia Lin | University of Washington | Large Scale Multiparty Silent Preprocessing for MPC from LPN |
| Hamed Nemati | KTH Royal Institute of Technology | Trustworthy Automatic Verification of Side-Channel Countermeasures for Binary Cryptographic Programs using the HoIBA libary |
| Karl Palmskog | KTH Royal Institute of Technology | Trustworthy Automatic Verification of Side-Channel Countermeasures for Binary Cryptographic Programs using the HoIBA libary |
| Chris Peikert | University of Michigan, Ann Arbor | Practical Third-Generation FHE and Bootstrapping |
| Dimitrios Skarlatos | Carnegie Mellon University | Scale-Out FHE LLMs on GPUs |
| Vinod Vaikuntanathan | Massachusetts Institute of Technology | Can Quantum Computers (Really) Factor? |
| Daniel Wichs | Northeastern University | Obfuscation, Proof Systems, and Secure Computation: A Research Program on Cryptography at the Simons Institute for the Theory of Computing |
| David Wu | University Of Texas At Austin | Fast Private Information Retrieval and More using Homomorphic Encryption |
Sustainability
| Recipient | University | Research title |
| Meeyoung Cha | Max Planck Institute | Forest-Blossom (Flossom): A New Framework for Sustaining Forest Biodiversity Through Outcome-Driven Remote Sensing Monitoring |
| Jingrui He | University of Illinois at Urbana–Champaign | Foundation Model Enabled Earth’s Ecosystem Monitoring |
| Pedro Lopes | University of Chicago | AI-powered Tools that Enable Engineers to Make & Re-make Sustainable Hardware |
| Cheng Yaw Low | Max Planck Institute | Forest-Blossom (Flossom): A New Framework for Sustaining Forest Biodiversity Through Outcome-Driven Remote Sensing Monitoring |
AI safety is a priority at Amazon. Our investment in safe, transparent, and responsible AI (RAI) includes collaboration with the global community and policymakers. We are members of and collaborate with organizations such as the Frontier Model Forum, the Partnership on AI, and other forums organized by government agencies such as the National Institute of Standards and Technology (NIST). Consistent with Amazon’s endorsement of the Korea Frontier AI Safety Commitments, we published our Frontier Model Safety Framework earlier this year.
During the development of the Nova Premier model, we conducted a comprehensive evaluation to assess its performance and safety. This included testing on both internal and public benchmarks and internal/automated and third-party red-teaming exercises. Once the final model was ready, we prioritized obtaining unbiased, third-party evaluations of the model’s robustness against RAI controls. In this post, we outline the key findings from these evaluations, demonstrating the strength of our testing approach and Amazon Premier’s standing as a safe model. Specifically, we cover our evaluations with two third-party evaluators: PRISM AI and ActiveFence.
Evaluation of Nova Premier against PRISM AI
PRISM Eval’s Behavior Elicitation Tool (BET) dynamically and systematically stress-tests AI models’ safety guardrails. The methodology focuses on measuring how many adversarial attempts (steps) it takes to get a model to generate harmful content across several key risk dimensions. The central metric is “steps to elicit” — the number of increasingly sophisticated prompting attempts required before a model generates an inappropriate response. A higher number of steps indicates stronger safety measures, as the model is more resistant to manipulation. The PRISM risk dimensions (inspired by the MLCommons AI Safety Benchmarks) include CBRNE weapons, violent crimes, non-violent crimes, defamation, and hate, amongst several others.
Using the BET Eval tool and its V1.0 metric, which is tailored toward non-reasoning models, we compared the recently released Nova models (Pro and Premier) to the latest models in the same class: Claude (3.5 v2 and 3.7 non-reasoning) and Llama4 Maverick, all available through Amazon Bedrock. PRISM BET conducts black-box evaluations (where model developers don’t have access to the test prompts) of models integrated with their API. The evaluation conducted with BET Eval MAX, PRISM’s most comprehensive/aggressive testing suite, revealed significant variations in safety against malicious instructions. Nova models demonstrated superior overall safety performance, with an average of 43 steps for Premier and 52 steps for Pro, compared to 37.7 for Claude 3.5 v2 and fewer than 12 steps for other models in the comparison set (namely, 9.9 for Claude3.7, 11.5 for Claude 3.7 thinking, and 6.5 for Maverick). This higher step count suggests that on average, Nova’s safety guardrails are more sophisticated and harder to circumvent through adversarial prompting. The figure below presents the number of steps per harm category evaluated through BET Eval MAX.
The PRISM evaluation provides valuable insights into the relative safety of different Amazon Bedrock models. Nova’s strong performance, particularly in hate speech and defamation resistance, represents meaningful progress in AI safety. However, the results also highlight the ongoing challenge of building truly robust safety measures into AI systems. As the field continues to evolve, frameworks like BET will play an increasingly important role in benchmarking and improving AI safety. As a part of this collaboration Nicolas Miailhe, CEO of PRISM Eval, said, “It’s incredibly rewarding for us to see Nova outperforming strong baselines using the BET Eval MAX; our aim is to build a long-term partnership toward safer-by-design models and to make BET available to various model providers.” Organizations deploying AI systems should carefully consider these safety metrics when selecting models for their applications.
Manual red teaming with ActiveFence
The AI safety & security company ActiveFence benchmarked Nova Premier on Bedrock on prompts distributed across Amazon’s eight core RAI categories. ActiveFence also evaluated Claude 3.7 (non-reasoning mode) and GPT 4.1 API on the same set. The flag rate on Nova Premier was lower than that on the other two models, indicating that Nova Premier is the safest of the three.
| Model | 3P Flag Rate [↓ is better] |
| Nova Premier | 12.0% |
| Sonnet 3.7 (non-reasoning) | 20.6% |
| GPT4.1 API | 22.4% |
“Our role is to think like an adversary but act in service of safety,” said Guy Paltieli from ActiveFence. “By conducting a blind stress test of Nova Premier under realistic threat scenarios, we helped evaluate its security posture in support of Amazon’s broader responsible-AI goals, ensuring the model could be deployed with greater confidence.”
These evaluations conducted with PRISM and ActiveFence give us confidence in the strength of our guardrails and our ability to protect our customers’ safety when they use our models. While these evaluations demonstrate strong safety performance, we recognize that AI safety is an ongoing challenge requiring continuous improvement. These assessments represent a point-in-time snapshot, and we remain committed to regular testing and enhancement of our safety measures. No AI system can guarantee perfect safety in all scenarios, which is why we maintain monitoring and response systems after deployment.
Acknowledgments: Vincent Ponzo, Elyssa Vincent
-
Funding & Business7 days agoKayak and Expedia race to build AI travel agents that turn social posts into itineraries
-
Jobs & Careers6 days ago
Mumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Mergers & Acquisitions6 days ago
Donald Trump suggests US government review subsidies to Elon Musk’s companies
-
Funding & Business6 days ago
Rethinking Venture Capital’s Talent Pipeline
-
Jobs & Careers6 days ago
Why Agentic AI Isn’t Pure Hype (And What Skeptics Aren’t Seeing Yet)
-
Funding & Business4 days ago
Sakana AI’s TreeQuest: Deploy multi-model teams that outperform individual LLMs by 30%
-
Jobs & Careers6 days ago
Astrophel Aerospace Raises ₹6.84 Crore to Build Reusable Launch Vehicle
-
Funding & Business7 days agoFrom chatbots to collaborators: How AI agents are reshaping enterprise work
-
Tools & Platforms6 days ago
Winning with AI – A Playbook for Pest Control Business Leaders to Drive Growth
-
Jobs & Careers4 days ago
Ilya Sutskever Takes Over as CEO of Safe Superintelligence After Daniel Gross’s Exit
