#Eval/Benchmark

AICMU researchers released SMDD-Bench, a benchmark of 502 small-molecule drug design tasks that use RDKit, ADMET-AI, and Boltz-2 as feedback loops. The authors argue that long-horizon planning, exploration, and learning from imperfect feedback remain open problems beyond math and coding, and the benchmark is available in Prime Intellect's Environments Hub for training with prime-rl.

AIBaseten tested the MetaInfer skills-only approach by having Claude Code build an inference engine, VibeQwen, for Qwen-3.6-35B-A3B in NVFP4 on a single B200. On single-stream text, VibeQwen decoded 90% faster than a tuned vLLM 0.25.1 deployment (1,792 vs. 943 TPS) and cut time to first token from 28 ms to 12 ms, with a 71% throughput gain at concurrency 32. The author notes this was an outcome-focused run that allowed some numerically different outputs as long as accuracy stayed at or above the BF16 baseline.

Why it matters: The post tests a skills-only inference engine method on a real model and states the speed and accuracy constraints used, helping readers judge how far such automated optimization can be trusted.

AIDesign Arena says users can give the same prompt to two models and vote blind to compare their outputs. The post advises choosing a model based on what you are making and directs readers to designarena.ai.

AIDesign Arena reports that every Astra-generated game in its code sample included accessibility features such as screen reader labels and reduced motion. In 60 of those games, the model never mentioned accessibility in its reasoning, suggesting it added these features by default.

AIAstra includes a header, footer, or navigation in 92% of its games, compared with 16% for Opus. It spends about 22 seconds of each game's thinking on the page structure, while Opus spends about 2 seconds.

AIOpus does the math before it writes the game. In a free kick game it works out exactly how hard to kick the ball to land on target. It also drafts code in its thinking in 91% of summaries. Astra does that in about 1 in 5.

AIDesign Arena says users can give the same prompt to two models and vote blind to compare their outputs. The post advises choosing a model based on what you are making and directs readers to designarena.ai.

AIDesign Arena reports that every Astra-generated game in its code sample included accessibility features such as screen reader labels and reduced motion. In 60 of those games, the model never mentioned accessibility in its reasoning, suggesting it added these features by default.

AIOpus does the math before it writes the game. In a free kick game it works out exactly how hard to kick the ball to land on target. It also drafts code in its thinking in 91% of summaries. Astra does that in about 1 in 5.

AIAstra includes a header, footer, or navigation in 92% of its games, compared with 16% for Opus. It spends about 22 seconds of each game's thinking on the page structure, while Opus spends about 2 seconds.

AIPerplexity has released several open source projects, including the pplx-decider-v1-27b multimodal decision model, the pplx-embed-v2-context-9b-preview contextual embeddings model, and the Lily local inference engine for Apple silicon. The post also lists the 0.6B on-device PII-Tracer classifier with its PII-TRACE benchmark, the WANDR research agent benchmark, and the Numbat and Bumblebee security tools, and says more open source releases are coming soon.

AISGLang's update adds a /v1/score endpoint that returns scores for requested labels such as Yes/No or A/B/C, avoiding the label loss of generate with top-k logprobs. Its multi-item scoring computes shared context once and keeps each candidate isolated, with 16-candidate p95 on Qwen3-8B dropping from 54.1 ms (Generate) to 20.6 ms.

AIKilo Code is offering Ling 3.1 Flash for free until October 13, with the model served by Novita Labs. Ant Ling's background post describes the model as roughly 560B total parameters with about 25B active per token and a context window of up to 1M tokens. Ant Ling says it scores 1,673 Elo on GDPVal-AA v2.1, 75.16 on FrontierSWE, and 65.35 on HealthBench Professional, and plans to open-source it soon.

AIAi2 released AstaBrief 8B, an open-weights model that turns a research question and retrieved literature excerpts into a cited report, along with its training data. The model runs as Fast mode in Asta, averaging 51.1 seconds per report versus 178.5 seconds for Thinking mode, about 3.5x faster. The post also describes filtering synthetic training data by citation density and building DPO pairs judged by two models that agreed.

Why it matters: The post explains how supervised fine-tuning, preference data, and citation-density filtering were used to build a cited-report model, which is useful for teams training their own models.

AILiquid AI shared a Hugging Face guide to multi-harness reinforcement learning for coding agents, in which a proxy records the token ids and logprobs vLLM samples so training works without changing the harness. Per the quoted post, LFM2.5-2.6B rose from 42% to 54% after training across four harnesses at once, and imitation fine-tuning on 3,189 rollouts from Qwen3.8-27B plateaued at 47.5%, below both RL runs. The proxy, trainer, tasks, SFT data, training code and seven trained models are described as open.

AILucas Beyer calls SWE-sweep a useful new benchmark, where agents must find and fix bugs in a repo checked out at an earlier commit, scored against unit tests from real later bugfixes. He notes two limitations: a model may find valid bugs that don't match the tested ones, and the construction makes training on the test set easy. He advises not overemphasizing small ranking differences once models score highly.

AICool eval. Simply ask an LLM “Land or Water?” and give it a latitude and longitude coordinate as text. Ask 16,200 times, plot as image. The models know. From compressing the internet.

AIServiceNow CoreAI introduced AutoSynthData, which uses a target model's failures and a stronger teacher's successes to generate and validate new agent training tasks. In EnterpriseOps Gym experiments, the Hybrid domain produced 2,000 samples and raised Gemma-4-26B-A4B-it mean Pass@1 by 7.2 percentage points, while the ITSM domain produced 1,994 samples and raised it from 18.77% to 27.18%.

Why it matters: The post shows how failure analysis, teacher demonstrations, and verifier checks combine into a repeatable pipeline for generating targeted agent training data.

AICoreWeave's new RL rollouts service uses ModelExpress and Router in NVIDIA Dynamo to speed up model weight reloads during RL post-training with minimal downtime. Working with NVIDIA and You.com, CoreWeave achieved 15× faster model reloads than its baseline while post-training Nemotron 3.5 Lightning. The speedup addresses GPUs sitting idle while inference workers wait to load updated weights between training iterations.

AIApple Machine Learning Research reports that discrete diffusion steps match the training distribution only when simultaneously written token positions are conditionally independent given already-fixed tokens. The authors show that per-position distributions cannot determine such dependence, and on the synthetic ScanAndAdd task, confidence-ranked groups of two or more positions were dependent and produced a generated distribution 29 times the sampling-noise floor in total variation.

AILiveKit now supports Microsoft's MAI-Transcribe-2-Streaming, MAI-Voice-2.1, and MAI-Voice-2.1-Flash for building voice agents. MAI-Transcribe-2-Streaming debuts at #1 on the Artificial Analysis accuracy leaderboard. Developers can pair it with MAI-Voice-2.1-Flash to build efficient and expressive voice agents.

AIPerplexity open-sourced pplx-decider-27b, a state-of-the-art multimodal Decision Model, and offers it through a new Decisions API. Input tokens cost $0.04 per million and output tokens are free, with further price cuts promised in the coming days.

AIARC Prize has published full ARC-AGI results for Alibaba's Qwen3.8-27B, with the testing run hosted by Baseten. The post links the public leaderboard, the open-source benchmarking repository for reproducing the results, and the testing policy.

AIJev, a reranker from TypeSafe AI, improved Wrangle's main-app search by 30.64% in matches per 200 candidates while costing 75% of the previous reranking cost. The gains were measured on searches recruiters actually run, rather than standard 10-candidate benchmarks, and the full benchmark is slated for release soon.

AIIn a business risk monitoring test by @edwardirby, the Jev judge matched the report quality of an ordinary LLM judge while missing no investigations, versus the LLM missing 5 of 11. The LLM's threat scores also flip-flopped from 0.35 to 0.68 to 0.50 on the same threat, while Jev was 250x cheaper and 3-6x faster.

AIFrançois Chollet argues the key difference between base LLMs and modern LRMs is a shift from transductive answer prediction to inductive prediction of the program or reasoning chain behind an answer. He says this enables test-time induction and substantial fluid intelligence in LRMs, which he claims base LLMs largely lack. He cites ARC 1 results: base LLMs remain around 10-15%, while LRMs of the same size or smaller saturated the benchmark in 2025.

AIFireworks AI's Ember-1 produced a near-identical composition to Kimi K3 on the same JavaScript painting prompt, according to Sophia Yang. Ember-1 used 9x fewer thinking tokens and ran 4x faster at half the cost.

AIGuillermo Rauch argues that the future is verification engineering, spanning proofs, end-to-end tests, benchmarks, and linters. He expects some of these tests to be deterministic and others agentic, and he says the approach looks great. The quoted post introduces e2e, an open-source agentic testing framework that mixes deterministic and agentic APIs and runs locally or in CI.

AIOdyssey introduces PROWL-2, a recursive loop that delivers up to 91% relative gains over the StarCraft world-model baseline. The post also reports stronger robot coordination in simulation.

AIGoogle Gemma relays a StudentBench study reporting that AI tutors matched expert human tutors on immediate GRE learning gains. The author reports 2,383 students and a cost of 7 cents per AI tutor hour versus $75 for an expert human hour. The post also says the top AI tutor beat expert human tutors on average in 5 of 7 academic topics, and that the data and paper are publicly available.

AIGoodfire reports that its method catches more unsafe biological requests than frontier model safeguards while dramatically reducing false positives. The claim is illustrated by a plot in the first tweet of the thread.