Sijun Tan

Harman Singh, Xiuyu Li, Kusha Sareen et al. · berkeley

Test-time scaling for complex reasoning tasks shows that leveraging inference-time compute, by methods such as independently sampling and aggregating multiple solutions, results in significantly better task outcomes. However, a critical bottleneck is verification: sampling is only effective if correct solutions can be reliably identified among candidates. While existing approaches typically evaluate candidates independently via scalar scoring, we demonstrate that models are substantially stronger at pairwise self-verification. Leveraging this insight, we introduce $V_1$, a framework that unifies generation and verification through efficient pairwise ranking. $V_1$ comprises two components: $V_1$-Infer, an uncertainty-guided algorithm using a tournament-based ranking that dynamically allocates self-verification compute to candidate pairs whose relative correctness is most uncertain; and $V_1$-PairRL, an RL framework that jointly trains a single model as both generator and pairwise self-verifier, ensuring the verifier adapts to the generator's evolving distribution. On code generation (LiveCodeBench, CodeContests, SWE-Bench) and math reasoning (AIME, HMMT) benchmarks, $V_1$-Infer improves Pass@1 by up to $10%$ over pointwise verification and outperforms recent test-time scaling methods while being significantly more efficient. Furthermore, $V_1$-PairRL achieves $7$--$9%$ test-time scaling gains over standard RL and pointwise joint training, and improves base Pass@1 by up to 8.7% over standard RL in a code-generation setting.

Sijun Tan, Siyuan Zhuang, Kyle Montgomery et al.

LLM-based judges have emerged as a scalable alternative to human evaluation and are increasingly used to assess, compare, and improve models. However, the reliability of LLM-based judges themselves is rarely scrutinized. As LLMs become more advanced, their responses grow more sophisticated, requiring stronger judges to evaluate them. Existing benchmarks primarily focus on a judge's alignment with human preferences, but often fail to account for more challenging tasks where crowdsourced human preference is a poor indicator of factual and logical correctness. To address this, we propose a novel evaluation framework to objectively evaluate LLM-based judges. Based on this framework, we propose JudgeBench, a benchmark for evaluating LLM-based judges on challenging response pairs spanning knowledge, reasoning, math, and coding. JudgeBench leverages a novel pipeline for converting existing difficult datasets into challenging response pairs with preference labels reflecting objective correctness. Our comprehensive evaluation on a collection of prompted judges, fine-tuned judges, multi-agent judges, and reward models shows that JudgeBench poses a significantly greater challenge than previous benchmarks, with many strong models (e.g., GPT-4o) performing just slightly better than random guessing. Overall, JudgeBench offers a reliable platform for assessing increasingly advanced LLM-based judges. Data and code are available at https://github.com/ScalerLab/JudgeBench.

Fangrui Huang, Souhad Chbeir, Arpandeep Khatua et al.

Large language models (LLMs) are increasingly used for mental-health support; yet prevailing evaluation methods--fluency metrics, preference tests, and generic dialogue benchmarks--fail to capture the clinically critical dimensions of psychotherapy. We introduce THERAPYGYM, a framework that evaluates and improves therapy chatbots along two clinical pillars: fidelity and safety. Fidelity is measured using the Cognitive Therapy Rating Scale (CTRS), implemented as an automated pipeline that scores adherence to CBT techniques over multi-turn sessions. Safety is assessed using a multi-label annotation scheme, covering therapy-specific risks (e.g., failing to address harm or abuse). To mitigate bias and unreliability in LLM-based judges, we further release THERAPYJUDGEBENCH, a validation set of 116 dialogues with 1,270 expert ratings for auditing and calibration against licensed clinicians. THERAPYGYM also serves as a training harness: CTRS and safety-based rewards drive RL with configurable patient simulations spanning diverse symptom profiles. Models trained in THERAPYGYM improve on expert ratings, with average CTRS rising from 0.10 to 0.60 (and 0.16 to 0.59 under LLM judges). Our work enables scalable development of therapy chatbots that are faithful to evidence-based practice and safer in high-stakes use.

Sijun Tan, Xiuyu Li, Shishir Patil et al.

Processing long contexts remains a challenge for large language models (LLMs) due to the quadratic computational and memory overhead of the self-attention mechanism and the substantial KV cache sizes during generation. We propose LLoCO, a novel approach to address this problem by learning contexts offline through context compression and in-domain parameter-efficient finetuning with LoRA. Our method enables an LLM to create a concise representation of the original context and efficiently retrieve relevant information to answer questions accurately. Our approach extends the effective context window of a 4k token LLaMA2-7B model to handle up to 128k tokens. We evaluate our approach on several long-context question-answering datasets, demonstrating that LLoCO significantly outperforms in-context learning while using $30\times$ fewer tokens during inference. LLoCO achieves up to $7.62\times$ speed-up during inference and $11.52\times$ higher throughput during finetuning, substantially reduces the cost of long document question answering. This makes it a promising solution for efficient long context processing. Our code is publicly available on https://github.com/jeffreysijuntan/lloco.

Kyle Montgomery, Sijun Tan, Yuqi Chen et al.

Test-time scaling is a powerful strategy for boosting the performance of large language models on complex reasoning tasks. While state-of-the-art approaches often employ generative verifiers to select the best solution from a pool of candidates, this method incurs prohibitive computational costs, limiting its practicality. In this work, we shift the focus to a more budget-aware paradigm: discriminative verification. We conduct a thorough empirical analysis and demonstrate that while discriminative verifiers may underperform in isolation, combining them with self-consistency in a hybrid approach creates a powerful and efficient test-time scaling mechanism. Notably, under a fixed compute budget, this hybrid approach surpasses state-of-the-art generative verification by a significant margin: achieving up to 15.3\% higher accuracy on AIME2025. Our findings establish that for practical, real-world applications, budget-aware scaling with discriminative verifiers is not only a "free" upgrade over self-consistency, but also a more effective and efficient alternative to costly generative techniques. Code is available at https://github.com/wang-research-lab/verification.

Xiaoyuan Zhu, Yaowen Ye, Tianyi Qiu et al.

As API access becomes a primary interface to large language models (LLMs), users often interact with black-box systems that offer little transparency into the deployed model. To reduce costs or maliciously alter model behaviors, API providers may discreetly serve quantized or fine-tuned variants, which can degrade performance and compromise safety. Detecting such substitutions is difficult, as users lack access to model weights and, in most cases, even output logits. To tackle this problem, we propose a rank-based uniformity test that can verify the behavioral equality of a black-box LLM to a locally deployed authentic model. Our method is accurate, query-efficient, and avoids detectable query patterns, making it robust to adversarial providers that reroute or mix responses upon the detection of testing attempts. We evaluate the approach across diverse threat scenarios, including quantization, harmful fine-tuning, jailbreak prompts, and full model substitution, showing that it consistently achieves superior statistical power over prior methods under constrained query budgets.

Sijun Tan, Jibang Wu, Xiaohui Bei et al.

Peer review systems such as conference paper review often suffer from the issue of miscalibration. Previous works on peer review calibration usually only use the ordinal information or assume simplistic reviewer scoring functions such as linear functions. In practice, applications like academic conferences often rely on manual methods, such as open discussions, to mitigate miscalibration. It remains an important question to develop algorithms that can handle different types of miscalibrations based on available prior knowledge. In this paper, we propose a flexible framework, namely least square calibration (LSC), for selecting top candidates from peer ratings. Our framework provably performs perfect calibration from noiseless linear scoring functions under mild assumptions, yet also provides competitive calibration results when the scoring function is from broader classes beyond linear functions and with arbitrary noise. On our synthetic dataset, we empirically demonstrate that our algorithm consistently outperforms the baseline which select top papers based on the highest average ratings.

Qizhi Zhang, Sijun Tan, Lichun Li et al.

Secure multi-party computation enables multiple mutually distrusting parties to perform computations on data without revealing the data itself, and has become one of the core technologies behind privacy-preserving machine learning. In this work, we present several improved privacy-preserving protocols for both linear and non-linear layers in machine learning. For linear layers, we present an extended beaver triple protocol for bilinear maps that significantly reduces communication of convolution layer. For non-linear layers, we introduce novel protocols for computing the sigmoid and softmax function. Both functions are essential building blocks for machine learning training of classification tasks. Our protocols are both more scalable and robust than prior constructions, and improves runtime performance by 3-17x. Finally, we introduce Morse-STF, an end-to-end privacy-preserving system for machine learning training that leverages all these improved protocols. Our system achieves a 1.8x speedup on logistic regression and 3.9-4.9x speedup on convolutional neural networks compared to prior state-of-the-art systems.

Sijun Tan, Brian Knott, Yuan Tian et al.

We introduce CryptGPU, a system for privacy-preserving machine learning that implements all operations on the GPU (graphics processing unit). Just as GPUs played a pivotal role in the success of modern deep learning, they are also essential for realizing scalable privacy-preserving deep learning. In this work, we start by introducing a new interface to losslessly embed cryptographic operations over secret-shared values (in a discrete domain) into floating-point operations that can be processed by highly-optimized CUDA kernels for linear algebra. We then identify a sequence of "GPU-friendly" cryptographic protocols to enable privacy-preserving evaluation of both linear and non-linear operations on the GPU. Our microbenchmarks indicate that our private GPU-based convolution protocol is over 150x faster than the analogous CPU-based protocol; for non-linear operations like the ReLU activation function, our GPU-based protocol is around 10x faster than its CPU analog. With CryptGPU, we support private inference and private training on convolutional neural networks with over 60 million parameters as well as handle large datasets like ImageNet. Compared to the previous state-of-the-art, when considering large models and datasets, our protocols achieve a 2x to 8x improvement in private inference and a 6x to 36x improvement for private training. Our work not only showcases the viability of performing secure multiparty computation (MPC) entirely on the GPU to enable fast privacy-preserving machine learning, but also highlights the importance of designing new MPC primitives that can take full advantage of the GPU's computing capabilities.