Learn about the amazing group of interns who joined us at Princeton and San Jose campuses this summer. Their hard work, fresh perspectives, and dedication have truly made an impact across the board, from cutting-edge research projects to innovative software development initiatives.
Our HorizonWeaver software edits driving scenes with instruction-guided AI, adding traffic, changing weather, and generalizing to unseen roads, all while preserving the safety-critical details that keep autonomous vehicle testing honest.
Multimodal retrievers are essential for knowledge-based visual question answering, where they retrieve external evidence for image-question pairs. However, existing contrastive training methods typically treat all unmatched query-document pairs as equally informative negatives, which is problematic because many unmatched documents may still be semantically relevant or partially useful. We propose Bayesian Data Reweighting, a probabilistic framework that models query-document importance as latent variables and adaptively infers posterior weights to downweight likely false negatives. With closed-form posterior updates under conjugate priors and stochastic EM optimization, our method consistently improves retrieval accuracy across three retrievers and seven knowledge-based VQA benchmarks.
Verizon and NEC Laboratories America turned live telecom fiber into a city-wide sensing network. By repurposing existing optical infrastructure, the system detects vibrations along fiber routes, cutting cable damage incidents and improving response times. The approach also enables AI models that adapt to new, unseen routes without manual labeling.
Video generation models can look physically convincing while getting the physics completely wrong. PhyCo, new research from our Media Analytics department, introduces continuous, controllable physical properties to video AI, allowing practitioners to specify friction, bounce, and force.
In this paper, we study the connection between an LLM’s output distribution and the data used to train it. Specifically, we study the degree to which an LLM’s next-token distribution agrees with the empirical next-token distribution (ENTD) given the context in the training data. The ENTD is an appealing target because it is the unrestricted global minimizer of the next-token cross entropy loss used for pretraining, as well as an easily interpretable function of the pretraining corpus. We find that for a significant fraction of inputs, the LLM’s distribution agrees with the ENTD almost perfectly, and the average agreement increases with model scale and training compute. Nevertheless, there is a long tail of input sequences where the LLM and ENTD differ significantly, and we examine several possible sources of this discrepancy across the transformer architecture, training procedure, and finite-sample noise in the ENTD estimate itself. More broadly, we hope our findings will encourage more work on “data-centric mechanistic interpretability,” a complement to standard mechanistic interpretability that opens the black box of how model behaviors arise from the data, rather than how they are encoded in the learned weight
As large language models increasingly serve as autonomous coding agents, code documentation must be optimized for agent comprehension rather than human readability. We frame agent-oriented documentation generation as a black-box optimization problem over the documentation space, where quality is measured solely by downstream code correctness. A central challenge for conventional LLM refinement methods is output couplingprogram entities are interdependent, and refining the documentation of one entity can invalidate its callers, resulting in a persistent whack-a-mole phenomenon during inference-time scaling. We propose DocSearch, a dependency-guided bi-level search framework that systematically exploits test-time feedback. The outer level conducts a priority search over the program-entity dependency DAG, enforcing a callee-before-caller refinement order to prevent downstream interference. The inner level performs a beam search over documentation refinements, using diversified error message sampling from self-generated unit tests to better exploit diagnostic signals and escape local optima. We provide theoretical guarantees of monotonic progress, showing that our worthy condition prevents regression while enabling efficient exploration. On DevEval+, DocSearch achieves a 90.7% solve rate with GPT-4o, outperforming the strongest baseline by 32.6%. Cross-language experiments further demonstrate that optimized documentation transfers effectively to different target programming languages.
Mix-CLAP from NEC Laboratories America delivers near-Transformer accuracy for sound event classification at a fraction of the compute cost, using dual lightweight encoders and adaptive, noise-aware inference for real-world edge deployment.
Chain-of-Thought (CoT) reasoning has driven recent gains of large language models (LLMs) on reasoning-intensive tasks by externalizing intermediate steps. However, excessive or redundant reasoning so-called overthinking can increase inference costs and lead LLMs toward incorrect conclusions. In this paper, we present REFRAIN (REFlective-Redundancy for Adaptive INference), a training-free framework that adaptively determines when to stop reasoning to mitigate overthinking. REFRAIN integrates a two-stage stop discriminator to identify reflective yet redundant reasoning and a sliding-window Upper Confidence Bound (SW-UCB) multi-armed bandit controller to dynamically adjust stopping thresholds according to problem difficulty without supervision or fine-tuning. Across four representative benchmarks and two model families, REFRAIN re-duces token usage by 20-55% while maintaining or improving accuracy compared to standard CoT prompting. Extensive ablation and robustness analyses demonstrate its stability across models, scorers, and prompt variations. In summary, our findings highlight when-to-stop as a new and practical axis of test-time scaling enabling models to reason not just more, but just enough.
Large language models (LLMs) often produce incorrect or outdated content after being employed. Efficient and accurate knowledge up-dates without costly retraining are a major challenge. This problem is particularly challenging in lifelong settings, where complex, unstructured knowledge must coexist without interference. We introduce RILKE (Representation Intervention for Lifelong KnowledgE Control), a robust and scalable method that treats knowledge control as interventions within the models representation space. Leveraging representation-space expressiveness, we identify two key properties enabling RILKE to achieve fine-grained control over complex, un-structured knowledge while maintaining general utility with frozen base weights. During training, RILKE learns paraphrase-robust and edit-localized modules that limit each update to a low-dimensional subspace to minimize cross-edit interference. At inference, a query-adaptive router selects the appropriate module to guide the models generation. Across LLaMA and Qwen models, RILKE scales effectively to large-scale benchmarks, demonstrating high edit success and strong paraphrase generalization while preserving general utility with modest memory overhead. These results show RILKE is an effective and scalable solution for lifelong knowledge control in LLMs.
Princeton Office
4 Independence Way, Suite 200
Princeton, NJ 08540
+1 609-520-1555
San Jose Office
2033 Gateway Place, Suite 200
San Jose, CA 95110
+1 408-863-6007
NEC Laboratories America, Inc. (NEC Labs) is the US-based center for NEC Corporation’s global network of corporate research laboratories. Our diverse research groups collaborate with industry, academia and governments to provide disruptive solutions to complex problems. A leader in the integration of IT and network technologies with more than 100 years of expertise, NEC provides a combination of products and solutions that cross-utilize the company’s experience and global resources to meet the complex and ever-changing needs of its customers.