arXiv AI September 30, 2026 By Dushyant Rajput (AltSlate Labs LLP), Nirdesh Chauhan (AltSlate Labs LLP), Siddharth Kosaraju (AltSlate Labs LLP)

Can a Cacheable Decision Model Follow Rules?

Read the original on arXiv AI →

The paper evaluates Certo, a small non‑generative decision model that scores candidate actions based on text. It compares a joint scorer that processes state, rules, and candidates together with a cacheable encoder that pre‑encodes candidates to reduce cost. Experiments show the cacheable approach loses rule sensitivity, while targeted counterfactual supervision can recover performance on synthetic tasks; however, on real rules the joint scorer still outperforms the cacheable version, and cross‑domain mixtures do not improve accuracy.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv Machine Learning

Jul 8

More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges

arXiv:2607. 05904v1 Announce Type: new Abstract: Training a language model against its own reference-free judgments (the premise of self-rewarding, self-play, and LLM-as-a-judge pipelines) assumes a model's verdict on a shown answer tracks correctness.

By Chenyu Zhou

llms reinforcement-learning

arXiv AI

3d ago

PACT: Pairwise-Anchored Calibrated Tuning for Single-Token Typed Decisions

The paper introduces PACT, a tuning method for single-token typed-decision models that leverages contrastive pair data to add four training terms—difference-in-differences margin, permutation-consistency, evidence-necessity, and ordinal transport cost—without requiring new annotations. PACT achieves comparable accuracy to existing recipes while reducing position bias and ordinal error, and it improves robustness and stability across seeds. The authors provide code, data splits, and trained adapters for reproducibility.

By Yida Lin

safety

arXiv AI

3d ago

Evaluating and Benchmarking the System One Model Jev

arXiv:2609.37647v1 Announce Type: cross Abstract: Jev is a commercial System One model from TypeSafe AI that does not generate text: given a state and typed questions, it returns a choice from fixed...

By Tobias Deu{\ss}er, Lorenz Sparrenberg, Rafet Sifa

benchmarks

arXiv Machine Learning

Aug 13

LODESTAR: Trustworthy Entropy Is Navigated, Not Merely Measured -- Reinforced Polarizer Keeps a Frozen LLM from Being Confidently Misled by the Wrong Evidence

arXiv:2608. 11922v1 Announce Type: cross Abstract: Predictive-distribution entropy makes a strong selection rule in retrieval-augmented question answering: across five QA benchmarks, keeping the candidate answer that a frozen respondent LLM produces with the lowest answer-token entropy lifts mean answer $F_1$ from 0.

By Po-Jen Ko, Che-Cheng Wu, Hung-Chun Hsu, Li-Yang Chang, Chuan-Ju Wang

llms rag nlp reinforcement-learning benchmarks

arXiv AI

Aug 18

Measuring Reward Hacking and Reasoning-Answer Decoupling Under Position-Confounded Optimization

arXiv:2608. 15445v1 Announce Type: new Abstract: When a reward is correct on every training example yet consistent with more than one goal, a model can acquire an unintended one, a failure known as goal misgeneralization.

By Suyash Maniyar, Armaan Sandhu, Abhishek Mishra

llms benchmarks

arXiv AI

Jul 7

Risk-Constrained Freshness-Aware Semantic Caching for Open-Web Retrieval-Augmented LLMs

arXiv:2607. 04281v1 Announce Type: cross Abstract: Semantic caching reduces the latency and cost of retrieval-augmented generation (RAG) by serving cached answers to semantically similar queries, but most existing methods do not model the time-varying freshness of open-web evidence.

By Muhammad Mansoor, Tahir Ahmad, Yeo-Chan Yoon

llms rag benchmarks

One email a morning, machine-written

Related stories

More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges

PACT: Pairwise-Anchored Calibrated Tuning for Single-Token Typed Decisions

Evaluating and Benchmarking the System One Model Jev

LODESTAR: Trustworthy Entropy Is Navigated, Not Merely Measured -- Reinforced Polarizer Keeps a Frozen LLM from Being Confidently Misled by the Wrong Evidence

Measuring Reward Hacking and Reasoning-Answer Decoupling Under Position-Confounded Optimization

Risk-Constrained Freshness-Aware Semantic Caching for Open-Web Retrieval-Augmented LLMs