Real-world LLM deployments increasingly rely on runtime-injected prohibitions--enterprise policies, PII redlines, tool boundaries--that vary per request and per tenant. Conventional post-training is structurally ill-suited: SFT hides the violation signal in compliant labels, and DPO's sequence-level preferences mismatch token-localized violations. We propose DUET, a token-selective on-policy distillation method for prohibition compliance. DUET pairs a teacher that sees the prohibition (positive) with an identical-weight teacher that does not (negative). Because the two teachers differ only in prohibition visibility, their per-token disagreement isolates the prohibition's causal effect--yielding a clean supervision signal uncontaminated by model capacity or mismatch. This disagreement drives two complementary mechanisms: signal cleaning, which discards agreement tokens as redundant or prefix-corrupted, and preference-directed learning, which pushes the student away from the negative teacher and toward the positive one at token granularity, embedding DPO-style optimization directly into OPD without offline preference data. We construct an industrial Prohibition-Compliance benchmark spanning five task families covering explicit-refusal, paraphrase robustness, and over-refusal. Across 1.5B-8B Qwen variants, DUET achieves 72.3-85.2% violation compliance while preserving 88-93% normal utility, dramatically outperforming teacher model and other distillation baselines. External evaluation on SysBench confirms improved safety alignment with minimal degradation on GSM8K and MATH-500.
Inference-time alignment methods, such as Best-of-$N$, offer a flexible alternative to training-based alignment by using reward models to select high-quality responses generated by a reference LLM. However, the efficacy of these methods is inherently limited by the response quality: if the reference LLM assigns negligible probability to high-reward responses, no selection strategy will succeed in finding aligned outputs. In this work, we propose Best-of-Better-$N$ (BoBN), an in context learning-based generation framework to address this challenge. Our method utilizes retrieval from high-reward examples relevant to the input query and task. Crucially, we introduce a restyling step where retrieved responses are rewritten by the reference LLM to align with the target task's format and style. These restyled examples are used in-context to shift the sampling distribution toward the high-reward region. We analytically characterize how in-context learning shifts the output distribution of pretrained transformers toward the high-reward region, resulting in provable benefits on the target task. We then evaluate BoBN on safety alignment and mathematical reasoning benchmarks across several reference LLMs. BoBN's higher-quality responses enable better performance to be achieved when the number of responses $N$ is fixed, and smaller $N$ required to achieve a target performance.
Khan Raiyan Ibne Reza, Sumaiya Tabassum Nimi, Omar Ibne Shahidcs.LG
We introduce KrishokChat, an 85,979-instance Bengali agricultural benchmark built from 284 government publications, 13 institutions, and six regional dialects. The benchmark comprises four tracks: General Knowledge QA, Treatment QA, Safety Refusal and Re-query, and Table QA. It also includes a 1,000-query Real-World Farmer Benchmark collected independently from field interviews to measure transfer to authentic farmer language. Every extracted instance retains provenance at the citation level. Treatment QA also provides a structured chemical-trace array for dosage-level auditability. We evaluated five zero-shot baselines and one fine-tuned model. Closed-book knowledge proves insufficient regardless of model scale. Oracle evidence narrows the gap, but leaves a persistent floor of 4.05 to 7.00% of chemical hallucinations. Fine-tuning on KrishokChat substantially outperforms the strongest zero-shot baseline on General QA Token F1. However, structured table reasoning and farmer-language transfer remain largely unsolved. An exploratory analysis further reveals that fine-tuning reduces refusal behavior on safety-critical queries, indicating that additional safety alignment is required beyond supervised fine-tuning. We release KrishokChat as a traceable benchmark and audit resource to support grounded and safety-aware agricultural language modeling for Bengali.
As Large Language Models (LLMs) advance toward open-ended autonomous agents, the mechanisms used to evaluate and guide their behavior must evolve accordingly. This work introduces the rubric as a unifying framework capturing this evolution, characterizing rubrics as a dynamic response to successive LLM paradigm shifts that recurs across otherwise independent efforts in evaluation, reinforcement learning, and safety alignment. We define rubrics as explicit criteria sets that transform complex quality judgments into structured and actionable standards, and demonstrate that their recurrence across these research threads is not coincidental. We systematically organize existing rubric designs, examine their construction and optimization, and analyze their role across evaluation and training. Rubrics manifest at three progressively deeper levels: at the evaluative level, they decompose holistic judgments into verifiable dimensions; at the training level, they serve as dense feedback signals providing process-level guidance where scalar rewards fall short; at the intrinsic level, they emerge dynamically from model behaviors, driving self-improvement. We further assess rubric reliability across generation quality, execution fidelity, theoretical constraints, and security threats, before surveying rubric-based benchmarks across diverse domains. By rendering assessment transparent and decomposable, rubrics translate human value expectations into machine-learnable signals, serving as the enduring bridge between human intentions and machine behavior.