Agentic loops, decision models, and test-time computation are reshaping how LLMs solve complex tasks and evaluate progress. Executable benchmarks and self-refinement cycles are replacing static leaderboards as models learn to revise their own outputs.
An executable benchmark for Chinese chess shows that LLM agents can point to the right move and still lose. Across 119 tactical endgames, 8,568 multi-turn trajectories and 12 frontier LLMs, closed-loop success collapses compared with static evaluatio...
A paired evaluation across 11 decision points for LLM agents finds hosted Jev outperforms open-weight Laya on 9 tasks, but neither beats chance on zero-shot routing. The pipeline audit turns a claimed 23.9% saving into an actual 4.3%.
Observational world models are not enough for modular LLM agents: a sequence does not say whether payment authorizes shipment or a hidden factor explains both. The FedCausalCompose framework shows that interventional error decreases with intervention...
An audit shows that LLM agents concentrate retrieval on a small core and leave rare states in a long tail, where prediction errors build up. The CTWM controller allocates prompt budget with a single exponent and reduces tokens by 24.48% on LongMemEva...
A benchmark on 824 multi-hop questions from FRAMES compares 18 RAG pipelines with an agent loop. The best pipeline reaches 78.9%, the agent 92.7%. The most awkward result concerns reranking: a small model drops accuracy by 9 points, a large one barel...
A runtime executes LLM agent tool calls out of trajectory order but commits only after validation. Speedups up to 1.35x on benchmarks with slow tools, zero false accepts across 4,010 records. A structural shift for self-hosted deployments focused on ...
AREX-2 explores self-improvement in LLM agents through reflection and long-horizon iterations. Trained on synthetic trajectories from machine learning and algorithmic programming tasks, built on Qwen3.8-27B, it achieves 81.8 on MLE-bench Lite and 70....
BAAI has released AREX-2, a 27B parameter agent model compatible with Qwen3.8. Its propose-measure-reflect-revise loop uses scores, logs, and errors to refine answers. With a 262,144 token context and GGUF files, it matters for self-hosted deployment...
Nemotron 3's IOI 2026 score is not just a milestone. The stated result rests on GenCorrect, a test-time compute loop that multiplies tokens per query. For on-premise deployments, open weights and NVFP4 quantization lower the hardware barrier, but ope...
NVIDIA's model card for Nemotron-Labs-3-Competitive-Coding describes a 550B-A55B model fine-tuned on 477,642 synthetic reasoning traces distilled from GLM-5.2. Combined with the GenCorrect strategy, it scored 535.4 out of 600 on the IOI 2026 problem ...
A consultant built ImaJev-4b, a 4-billion-parameter model based on Qwen3.5-4B with LoRA and a small decision head. After an initial fine-tuning that degraded reasoning, careful data curation led to first place on JevBench among 91 models and third on...
