GenAC introduces generative critics with chain-of-thought reasoning and in-context conditioning to improve value approximation and downstream RL performance in LLMs compared to value-based and value-free baselines.
Title resolution pending
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
dataset 1
citation-polarity summary
years
2026 8roles
dataset 1polarities
use dataset 1representative citing papers
DeepWeb-Bench is a benchmark requiring massive cross-source evidence collection and long-horizon derivation, with evaluations on nine frontier models showing derivation and calibration as primary failure modes.
Higher worker capability in multi-agent LLM systems increases semantic hijacking attack success rates via linguistic certainty in reports, with heterogeneous ensembles reducing ASR from 52.8% to 2.0%.
LAQuant improves long-decoding accuracy on quantized reasoning models like Qwen3-4B by 15pp on AIME25 via layer-wise lookahead loss, achieving 3.42x speedup over FP16.
Hint Tuning reduces token usage 24-66% (31.5% avg) in reasoning models via 1K self-annotated samples aligned to an instruct model's capabilities while keeping benchmark accuracy.
Correcting DeepSpeed optimizer and OpenRLHF loss bugs reveals SFT-then-RL outperforms mixed-policy methods by 3.8-22.2 points on math benchmarks.
SnapMLA achieves up to 1.91x higher throughput in long-output MLA decoding using FP8 quantization and specialized kernels while keeping benchmark quality near the BF16 baseline.
A summary of expert opinions on AI's impact on the research-education environment in physics from a KITP discussion session.
citing papers explorer
-
Bringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning
GenAC introduces generative critics with chain-of-thought reasoning and in-context conditioning to improve value approximation and downstream RL performance in LLMs compared to value-based and value-free baselines.
-
DeepWeb-Bench: A Deep Research Benchmark Demanding Massive Cross-Source Evidence and Long-Horizon Derivation
DeepWeb-Bench is a benchmark requiring massive cross-source evidence collection and long-horizon derivation, with evaluations on nine frontier models showing derivation and calibration as primary failure modes.
-
The Capability Paradox: How Smarter Auditors Make Multi-Agent Systems Less Secure
Higher worker capability in multi-agent LLM systems increases semantic hijacking attack success rates via linguistic certainty in reports, with heterogeneous ensembles reducing ASR from 52.8% to 2.0%.
-
LAQuant: A Simple Overhead-free Large Reasoning Model Quantization by Layer-wise Lookahead Loss
LAQuant improves long-decoding accuracy on quantized reasoning models like Qwen3-4B by 15pp on AIME25 via layer-wise lookahead loss, achieving 3.42x speedup over FP16.
-
Hint Tuning: Less Data Makes Better Reasoners
Hint Tuning reduces token usage 24-66% (31.5% avg) in reasoning models via 1K self-annotated samples aligned to an instruct model's capabilities while keeping benchmark accuracy.
-
SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning
Correcting DeepSpeed optimizer and OpenRLHF loss bugs reveals SFT-then-RL outperforms mixed-policy methods by 3.8-22.2 points on math benchmarks.
-
SnapMLA: Efficient Long-Context MLA Decoding via Hardware-Aware FP8 Quantized Pipelining
SnapMLA achieves up to 1.91x higher throughput in long-output MLA decoding using FP8 quantization and specialized kernels while keeping benchmark quality near the BF16 baseline.
-
AI and the Research-Education Environment of Physics
A summary of expert opinions on AI's impact on the research-education environment in physics from a KITP discussion session.