Be part of our every day and weekly newsletters for the newest updates and unique content material on industry-leading AI protection. Be taught Extra
Enterprises must know if the fashions that energy their purposes and brokers work in real-life situations. This sort of analysis can generally be complicated as a result of it’s arduous to foretell particular situations. A revamped model of the RewardBench benchmark appears to provide organizations a greater thought of a mannequin’s real-life efficiency.
The Allen Institute of AI (Ai2) launched RewardBench 2, an up to date model of its reward mannequin benchmark, RewardBench, which they declare supplies a extra holistic view of mannequin efficiency and assesses how fashions align with an enterprise’s objectives and requirements.
Ai2 constructed RewardBench with classification duties that measure correlations by inference-time compute and downstream coaching. RewardBench primarily offers with reward fashions (RM), which may act as judges and consider LLM outputs. RMs assign a rating or a “reward” that guides reinforcement studying with human suggestions (RHLF).
Nathan Lambert, a senior analysis scientist at Ai2, advised VentureBeat that the primary RewardBench labored as meant when it was launched. Nonetheless, the mannequin setting quickly advanced, and so ought to its benchmarks.
“As reward fashions grew to become extra superior and use instances extra nuanced, we shortly acknowledged with the neighborhood that the primary model didn’t absolutely seize the complexity of real-world human preferences,” he stated.
Lambert added that with RewardBench 2, “we got down to enhance each the breadth and depth of analysis—incorporating extra numerous, difficult prompts and refining the methodology to replicate higher how people really decide AI outputs in observe.” He stated the second model makes use of unseen human prompts, has a tougher scoring setup and new domains.
Utilizing evaluations for fashions that consider
Whereas reward fashions check how nicely fashions work, it’s additionally vital that RMs align with firm values; in any other case, the fine-tuning and reinforcement studying course of can reinforce dangerous habits, resembling hallucinations, cut back generalization, and rating dangerous responses too excessive.
RewardBench 2 covers six completely different domains: factuality, exact instruction following, math, security, focus and ties.
“Enterprises ought to use RewardBench 2 in two other ways relying on their software. In the event that they’re performing RLHF themselves, they need to undertake one of the best practices and datasets from main fashions in their very own pipelines as a result of reward fashions want on-policy coaching recipes (i.e. reward fashions that mirror the mannequin they’re making an attempt to coach with RL). For inference time scaling or information filtering, RewardBench 2 has proven that they’ll choose one of the best mannequin for his or her area and see correlated efficiency,” Lambert stated.
Lambert famous that benchmarks like RewardBench provide customers a technique to consider the fashions they’re selecting primarily based on the “dimensions that matter most to them, fairly than counting on a slender one-size-fits-all rating.” He stated the concept of efficiency, which many analysis strategies declare to evaluate, could be very subjective as a result of a very good response from a mannequin extremely is dependent upon the context and objectives of the consumer. On the identical time, human preferences get very nuanced.
Ai 2 launched the primary model of RewardBench in March 2024. On the time, the corporate stated it was the primary benchmark and leaderboard for reward fashions. Since then, a number of strategies for benchmarking and bettering RM have emerged. Researchers at Meta’s FAIR got here out with reWordBench. DeepSeek launched a brand new approach known as Self-Principled Critique Tuning for smarter and scalable RM.
How fashions carried out
Since RewardBench 2 is an up to date model of RewardBench, Ai2 examined each present and newly skilled fashions to see in the event that they proceed to rank excessive. These included a wide range of fashions, resembling variations of Gemini, Claude, GPT-4.1, and Llama-3.1, together with datasets and fashions like Qwen, Skywork, and its personal Tulu.
The corporate discovered that bigger reward fashions carry out greatest on the benchmark as a result of their base fashions are stronger. Total, the strongest-performing fashions are variants of Llama-3.1 Instruct. By way of focus and security, Skywork information “is especially useful,” and Tulu did nicely on factuality.
Ai2 stated that whereas they consider RewardBench 2 “is a step ahead in broad, multi-domain accuracy-based analysis” for reward fashions, they cautioned that mannequin analysis must be primarily used as a information to select fashions that work greatest with an enterprise’s wants.
