Building Production-Ready Probes For Gemini

• URL: http://arxiv.org/abs/2601.11516v2
• Date: Mon, 19 Jan 2026 16:05:05 GMT
• Title: Building Production-Ready Probes For Gemini
• Abstract summary: We find that the shift from short-context to long-context inputs is difficult for existing probe architectures.<n>We propose several new probe architectures that handle this long-context distribution shift.<n>We find early positive results using AlphaEvolve to automate improvements in both probe architecture search and adaptive red teaming.
• Score: 21.610660065701243
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Frontier language model capabilities are improving rapidly. We thus need stronger mitigations against bad actors misusing increasingly powerful systems. Prior work has shown that activation probes may be a promising misuse mitigation technique, but we identify a key remaining challenge: probes fail to generalize under important production distribution shifts. In particular, we find that the shift from short-context to long-context inputs is difficult for existing probe architectures. We propose several new probe architectures that handle this long-context distribution shift. We evaluate these probes in the cyber-offensive domain, testing their robustness against various production-relevant distribution shifts, including multi-turn conversations, long context prompts, and adaptive red teaming. Our results demonstrate that while our novel architectures address context length, a combination of architecture choice and training on diverse distributions is required for broad generalization. Additionally, we show that pairing probes with prompted classifiers achieves optimal accuracy at a low cost due to the computational efficiency of probes. These findings have informed the successful deployment of misuse mitigation probes in user-facing instances of Gemini, Google's frontier language model. Finally, we find early positive results using AlphaEvolve to automate improvements in both probe architecture search and adaptive red teaming, showing that automating some AI safety research is already possible.

Related papers

• Apodex Discovery: Reality Benchmarks and Environments for Evaluating and Building Discoverative Artificial Intelligence [86.61561123749011]
  We introduce Apodex Discovery, a framework for building and evaluating discoverative AI through the heavy-duty solver.<n>First, a problem-scouting process surveyed 561 industries across 16 sectors, assembled 423 high-value real-world problems, and selected 20 for the initial release.<n>Second, a common environment-task-episode abstraction provides data, tools, constraints, feedback, trajectory recording, and verification of intermediate artifacts and final submissions.
  arXiv  Detail & Related papers  (2026-08-11T18:48:40Z)
• Rethinking AI-Generated Text Detection: A Strong Baseline and the Distribution-Shift Problem That Remains [27.078441680572254]
  We show that a plain, fully fine-tuned RoBERTa matches or exceeds the specialized detectors those benchmarks are built around.<n>The remaining challenge is the distribution shift.<n>We identify a key failure mode: under distribution shift, the detector can assign high-confidence machine labels to human-written text.
  arXiv  Detail & Related papers  (2026-07-04T03:20:23Z)
• GENESIS: Harnessing AI Agents for Autonomous 6G RAN Synthesis, Research, and Testing [13.436656977622691]
  Large Language Models (LLMs) have compressed comparable R&D work in general software engineering from days to minutes.<n>We present GENESIS, an agentic Artificial Intelligence (AI) framework that converts intents into solutions validated with over-the-air experiments.
  arXiv  Detail & Related papers  (2026-05-26T17:58:43Z)
• Linear Probe Accuracy Scales with Model Size and Benefits from Multi-Layer Ensembling [0.0]
  Linear probes can detect when language models produce outputs they "know" are wrong.<n>We show that combining probes from multiple layers into an ensemble recovers strong performance even where single-layer probes fail.
  arXiv  Detail & Related papers  (2026-04-15T01:21:52Z)
• Hidden Failures in Robustness: Why Supervised Uncertainty Quantification Needs Better Evaluation [40.069986955093526]
  Recent work has shown that the hidden states of large language models contain signals useful for uncertainty estimation and hallucination detection.<n>Yet it remains unclear how robust existing methods are, and which probe designs provide uncertainty estimates that are reliable under distribution shift.<n>We present a systematic study of supervised uncertainty probes across models, tasks, and OOD settings, training over 2,000 probes while varying the representation layer, feature type, and token aggregation strategy.
  arXiv  Detail & Related papers  (2026-04-13T16:11:33Z)
• Improving Coherence and Persistence in Agentic AI for System Optimization [9.443037059325086]
  Engram organizes exploration into a sequence of agents that iteratively design, test, and analyze mechanisms.<n>We find that Engram exhibits superior performance across diverse domains including multi-cloud multicast inference, LLM request routing, and optimizing KV cache reuse in databases with natural language queries.
  arXiv  Detail & Related papers  (2026-03-22T17:04:50Z)
• Scaling Agentic Verifier for Competitive Coding [66.11758166379092]
  Large language models (LLMs) have demonstrated strong coding capabilities but still struggle to solve competitive programming problems correctly in a single attempt.<n>Execution-based re-ranking offers a promising test-time scaling strategy, yet existing methods are constrained by either difficult test case generation or inefficient random input sampling.<n>We propose Agentic Verifier, an execution-based agent that actively reasons about program behaviors and searches for highly discriminative test inputs.
  arXiv  Detail & Related papers  (2026-02-04T06:30:40Z)
• GenPilot: A Multi-Agent System for Test-Time Prompt Optimization in Image Generation [13.197958581564256]
  We propose a test-time prompt optimization strategy that operates directly on the input text.<n>Our approach is model-agnostic, interpretable, and well-suited for handling long and complex prompts.
  arXiv  Detail & Related papers  (2025-10-08T16:51:52Z)
• UltraHorizon: Benchmarking Agent Capabilities in Ultra Long-Horizon Scenarios [63.67884284105684]
  We introduce textbfUltraHorizon, a novel benchmark that measures the foundational capabilities essential for complex real-world challenges.<n>Agents are designed in long-horizon discovery tasks where they must iteratively uncover hidden rules.<n>Our experiments reveal that LLM-agents consistently underperform in these settings, whereas human participants achieve higher scores.
  arXiv  Detail & Related papers  (2025-09-26T02:04:00Z)
• Detecting High-Stakes Interactions with Activation Probes [10.421494494629421]
  This paper examines activation probes for detecting "high-stakes" interactions, where the text indicates that the interaction might lead to significant harm.<n>We evaluate several probe architectures trained on synthetic data, and find them to exhibit robust generalization to diverse, out-of-distribution, real-world data.<n>Our experiments also highlight the potential of building resource-aware hierarchical monitoring systems.
  arXiv  Detail & Related papers  (2025-06-12T15:20:33Z)
• When Disagreements Elicit Robustness: Investigating Self-Repair Capabilities under LLM Multi-Agent Disagreements [56.29265568399648]
  We argue that disagreements prevent premature consensus and expand the explored solution space.<n>Disagreements on task-critical steps can derail collaboration depending on the topology of solution paths.
  arXiv  Detail & Related papers  (2025-02-21T02:24:43Z)
• A Self-Supervised Task for Fault Detection in Satellite Multivariate Time Series [45.31237646796715]
  This work proposes a novel approach leveraging Physics-Informed Real NVP neural networks, renowned for their ability to model complex and high-dimensional distributions.<n>The experiments involve various configurations, including pre-training with self-supervision, multi-task learning, and standalone self-supervised training.<n>Results indicate significant performance improvements across all settings.
  arXiv  Detail & Related papers  (2024-07-03T07:19:41Z)
• Beyond Two-Tower Matching: Learning Sparse Retrievable Cross-Interactions for Recommendation [80.19762472699814]
  Two-tower models are a prevalent matching framework for recommendation, which have been widely deployed in industrial applications. It suffers two main challenges, including limited feature interaction capability and reduced accuracy in online serving. We propose a new matching paradigm named SparCode, which supports not only sophisticated feature interactions but also efficient retrieval.
  arXiv  Detail & Related papers  (2023-11-30T03:13:36Z)
• Complexity Matters: Rethinking the Latent Space for Generative Modeling [65.64763873078114]
  In generative modeling, numerous successful approaches leverage a low-dimensional latent space, e.g., Stable Diffusion. In this study, we aim to shed light on this under-explored topic by rethinking the latent space from the perspective of model complexity.
  arXiv  Detail & Related papers  (2023-07-17T07:12:29Z)
• Searching for the Fakes: Efficient Neural Architecture Search for General Face Forgery Detection [4.19882227319634]
  We develop an end-to-end framework based on neural architecture search (NAS) for deepfake detection. We propose a novel performance estimation metric, which guides the search process to select more general models. Compared with state-of-the-art networks artificially designed, our method achieves competitive performance in both in-dataset and cross-dataset scenarios.
  arXiv  Detail & Related papers  (2023-06-15T03:01:13Z)
• PreTR: Spatio-Temporal Non-Autoregressive Trajectory Prediction Transformer [0.9786690381850356]
  We introduce a model called PRediction Transformer (PReTR) that extracts features from the multi-agent scenes by employing a factorized-temporal attention module. It shows less computational needs than previously studied models with empirically better results. We leverage encoder-decoder Transformer networks for parallel decoding a set of learned object queries.
  arXiv  Detail & Related papers  (2022-03-17T12:52:23Z)
• Low-Complexity Probing via Finding Subnetworks [101.43611844196064]
  We propose a subtractive-based probe for probing neural networks for linguistic properties. We find that the subnetwork probe achieves both higher accuracy on pre-trained models and lower accuracy on random models. By varying the complexity of each probe, we show that subnetwork probing-dominates probing in that it higher accuracy given any budget of probe complexity.
  arXiv  Detail & Related papers  (2021-04-08T05:11:21Z)
• Resource Allocation in Multi-armed Bandit Exploration: Overcoming Sublinear Scaling with Adaptive Parallelism [107.48538091418412]
  We study exploration in multi-armed bandits when we have access to a divisible resource that can be allocated in varying amounts to arm pulls. We focus in particular on the allocation of distributed computing resources, where we may obtain results faster by allocating more resources per pull.
  arXiv  Detail & Related papers  (2020-10-31T18:19:29Z)
• Recent Developments Combining Ensemble Smoother and Deep Generative Networks for Facies History Matching [58.720142291102135]
  This research project focuses on the use of autoencoders networks to construct a continuous parameterization for facies models. We benchmark seven different formulations, including VAE, generative adversarial network (GAN), Wasserstein GAN, variational auto-encoding GAN, principal component analysis (PCA) with cycle GAN, PCA with transfer style network and VAE with style loss.
  arXiv  Detail & Related papers  (2020-05-08T21:32:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.