https://benchmark-radar.org/blog/ One page per collection day: what new benchmarks and evaluations appeared, how strong the evidence was, and which claims the record does not support. en Sun, 11 Oct 2026 00:00:00 GMT https://benchmark-radar.org/blog/2026-10-11/ https://benchmark-radar.org/blog/2026-10-11/ Sun, 11 Oct 2026 00:00:00 GMT The new Two-Stage Measurement of Update Adherence benchmark tests whether a memory-enabled language-model agent follows a revised requirement rather than… https://benchmark-radar.org/blog/2026-10-10/ https://benchmark-radar.org/blog/2026-10-10/ Sat, 10 Oct 2026 00:00:00 GMT The new milipoint-runs artifact exposes split leakage in the MiliPoint radar benchmark: overlapping windows let most test examples share frames with… https://benchmark-radar.org/blog/2026-10-09/ https://benchmark-radar.org/blog/2026-10-09/ Fri, 09 Oct 2026 00:00:00 GMT TRACE is a new diagnostic protocol for checking whether an agent’s benchmark score changed because its behavior changed or because the evaluator changed… https://benchmark-radar.org/blog/2026-10-08/ https://benchmark-radar.org/blog/2026-10-08/ Thu, 08 Oct 2026 00:00:00 GMT A new runtime-verification study exposes a data-design gap in existing agent-safety benchmarks. It replayed AgentDojo and STAC trajectories through formal… https://benchmark-radar.org/blog/2026-10-07/ https://benchmark-radar.org/blog/2026-10-07/ Wed, 07 Oct 2026 00:00:00 GMT The new Agent Network Traffic benchmark (ANT) tests whether security teams can audit large language model agents through network traffic without… https://benchmark-radar.org/blog/2026-10-05/ https://benchmark-radar.org/blog/2026-10-05/ Mon, 05 Oct 2026 00:00:00 GMT The new “Passing the Test You Trained On” study evaluates 15 prompt-injection detectors on benign and injected outputs reconstructed from real agent… https://benchmark-radar.org/blog/2026-10-04/ https://benchmark-radar.org/blog/2026-10-04/ Sun, 04 Oct 2026 00:00:00 GMT The new 2026 Cryo-EM Benchmark Dataset provides 12 protein structures deposited after every test set disclosed for MICA, an artificial intelligence system… https://benchmark-radar.org/blog/2026-10-03/ https://benchmark-radar.org/blog/2026-10-03/ Sat, 03 Oct 2026 00:00:00 GMT The updated retinal benchmark audit provides image-level manifests for patient-grouped, source-matched, and leave-one-source-out evaluations, plus… https://benchmark-radar.org/blog/2026-10-02/ https://benchmark-radar.org/blog/2026-10-02/ Fri, 02 Oct 2026 00:00:00 GMT Video-Index is a new meta-benchmark that audits whether video tests actually require video understanding. Its five-level “attack pyramid” tests… https://benchmark-radar.org/blog/2026-10-01/ https://benchmark-radar.org/blog/2026-10-01/ Thu, 01 Oct 2026 00:00:00 GMT The new study “Dating the Model” reports that hidden current-date text injected into system prompts changed measured performance across nine large… https://benchmark-radar.org/blog/2026-09-30/ https://benchmark-radar.org/blog/2026-09-30/ Wed, 30 Sep 2026 00:00:00 GMT The newly released UserProxyBench tests a hidden dependency in interactive agent benchmarks: whether the language model playing the user follows its… https://benchmark-radar.org/blog/2026-09-29/ https://benchmark-radar.org/blog/2026-09-29/ Tue, 29 Sep 2026 00:00:00 GMT TraceDance is a new release that turns real agent deployment traces into targeted benchmarks for user-specified undesirable behaviors. It evaluates the… https://benchmark-radar.org/blog/2026-09-28/ https://benchmark-radar.org/blog/2026-09-28/ Mon, 28 Sep 2026 00:00:00 GMT The new Finance Agents Benchmark packages financial due diligence as an end-to-end agent test: 50 tasks, 160 documents, 231 grading criteria, an execution… https://benchmark-radar.org/blog/2026-09-27/ https://benchmark-radar.org/blog/2026-09-27/ Sun, 27 Sep 2026 00:00:00 GMT The new Kalyvox Voice Benchmark 2026 evaluates one voice agent through 240 controlled French and English calls across 12 scenario families. It reports… https://benchmark-radar.org/blog/2026-09-26/ https://benchmark-radar.org/blog/2026-09-26/ Sat, 26 Sep 2026 00:00:00 GMT The new repo2bench release converts Python repositories into coding-agent tasks mined from version history or created by removing functions. It admits… https://benchmark-radar.org/blog/2026-09-25/ https://benchmark-radar.org/blog/2026-09-25/ Fri, 25 Sep 2026 00:00:00 GMT The new PartHackBench release tests whether partial-credit scorers for long-running tool agents reward misleading progress. It compares honest and… https://benchmark-radar.org/blog/2026-09-24/ https://benchmark-radar.org/blog/2026-09-24/ Thu, 24 Sep 2026 00:00:00 GMT The new Uncheatable Eval benchmark evaluates base language models by measuring how efficiently they losslessly compress newly published text that is… https://benchmark-radar.org/blog/2026-09-23/ https://benchmark-radar.org/blog/2026-09-23/ Wed, 23 Sep 2026 00:00:00 GMT A new B3DB audit tests whether stereoisomers inflate blood–brain barrier prediction scores because common two-dimensional molecular representations map… https://benchmark-radar.org/blog/2026-09-22/ https://benchmark-radar.org/blog/2026-09-22/ Tue, 22 Sep 2026 00:00:00 GMT The new Curation-Bench release evaluates whether coding agents can improve training-data selection through an iterative workflow: agents inspect data… https://benchmark-radar.org/blog/2026-09-20/ https://benchmark-radar.org/blog/2026-09-20/ Sun, 20 Sep 2026 00:00:00 GMT The newly released TC-Slang Guardrail Benchmark evaluates Traditional Chinese prompt-injection defenses by slang year and pairs each slang example with a… https://benchmark-radar.org/blog/2026-09-18/ https://benchmark-radar.org/blog/2026-09-18/ Fri, 18 Sep 2026 00:00:00 GMT The new SonoCorpus and SonoBase release pairs 456,963 ultrasound images and 1,626,085 expert segmentation masks with an interactive segmentation model… https://benchmark-radar.org/blog/2026-09-17/ https://benchmark-radar.org/blog/2026-09-17/ Thu, 17 Sep 2026 00:00:00 GMT The new lexEN benchmark replaces disputed word-sense labels with a conservative, human-adjudicated correction layer: 211 labels changed and 56 removed… https://benchmark-radar.org/blog/2026-09-16/ https://benchmark-radar.org/blog/2026-09-16/ Wed, 16 Sep 2026 00:00:00 GMT A new audit of 254 SWE-bench coding-agent submissions reports that the leaderboard cannot reliably order its leading entries: the top two resolve the same… https://benchmark-radar.org/blog/2026-09-15/ https://benchmark-radar.org/blog/2026-09-15/ Tue, 15 Sep 2026 00:00:00 GMT The new Sophea release evaluates a production Greek-English speech recognizer against nine simultaneous gates covering both languages, language… https://benchmark-radar.org/blog/2026-09-14/ https://benchmark-radar.org/blog/2026-09-14/ Mon, 14 Sep 2026 00:00:00 GMT The new KNOWS Benchmark evaluates web agents on 110 tasks that combine web research with producing documents, spreadsheets, and slide decks in Google… https://benchmark-radar.org/blog/2026-09-13/ https://benchmark-radar.org/blog/2026-09-13/ Sun, 13 Sep 2026 00:00:00 GMT The newly captured MusicAI background-music benchmark tests audio-capable large language models on 2,000 questions under clean audio, white noise, and 55… https://benchmark-radar.org/blog/2026-09-12/ https://benchmark-radar.org/blog/2026-09-12/ Sat, 12 Sep 2026 00:00:00 GMT The newly released Benchmark Radar combines a searchable catalog of AI evaluations with links to datasets and code, mentions in model cards and technical… https://benchmark-radar.org/blog/2026-09-11/ https://benchmark-radar.org/blog/2026-09-11/ Fri, 11 Sep 2026 00:00:00 GMT The new oncology visual question answering benchmark builds test items from private, single-institution radiology reports paired with three-dimensional… https://benchmark-radar.org/blog/2026-09-10/ https://benchmark-radar.org/blog/2026-09-10/ Thu, 10 Sep 2026 00:00:00 GMT The new “Double Measurement Confound” paper identifies a recurring agent-evaluation problem in this feed: a fixed software scaffold may make execution… https://benchmark-radar.org/blog/2026-09-09/ https://benchmark-radar.org/blog/2026-09-09/ Wed, 09 Sep 2026 00:00:00 GMT The new “Style Over Substance” study tests automated safety judges by preserving an answer byte-for-byte while adding tone-only wrappers, such as… https://benchmark-radar.org/blog/2026-09-08/ https://benchmark-radar.org/blog/2026-09-08/ Tue, 08 Sep 2026 00:00:00 GMT The new URL Intelligence Benchmark evaluates agents and web-analysis tools on operational hazards that ordinary answer-accuracy tests miss: redirects… https://benchmark-radar.org/blog/2026-09-07/ https://benchmark-radar.org/blog/2026-09-07/ Mon, 07 Sep 2026 00:00:00 GMT TruthInsightBench is a new benchmark for scientific-discovery agents that replaces reproduction-style tasks with 40 blind tasks from peer-reviewed… https://benchmark-radar.org/blog/2026-09-06/ https://benchmark-radar.org/blog/2026-09-06/ Sun, 06 Sep 2026 00:00:00 GMT The new Amharic Automatic Speech Recognition benchmark evaluates open models on 1,548 clips that its publisher says were unavailable for training. It… https://benchmark-radar.org/blog/2026-09-05/ https://benchmark-radar.org/blog/2026-09-05/ Sat, 05 Sep 2026 00:00:00 GMT Several new releases in this captured feed turn agent evaluation into domain-specific workflow testing. The clearest is… https://benchmark-radar.org/blog/2026-09-04/ https://benchmark-radar.org/blog/2026-09-04/ Fri, 04 Sep 2026 00:00:00 GMT A new preregistered audit, “Clean Engineering, Unstable Measurement,” tests whether a black-box large language model judge behaves like a stable… https://benchmark-radar.org/blog/2026-09-03/ https://benchmark-radar.org/blog/2026-09-03/ Thu, 03 Sep 2026 00:00:00 GMT Among today’s captured releases, EarlyEval introduces early outcome prediction: it estimates an agent’s final result from intermediate behavior and stops… https://benchmark-radar.org/blog/2026-09-02/ https://benchmark-radar.org/blog/2026-09-02/ Wed, 02 Sep 2026 00:00:00 GMT The newly released InSight benchmark tests agents that must actively interact with visualizations to verify 21,349 claims, rather than answer once from a… https://benchmark-radar.org/blog/2026-09-01/ https://benchmark-radar.org/blog/2026-09-01/ Tue, 01 Sep 2026 00:00:00 GMT EleutherAI released Language Model Evaluation Harness v0.4.13 with fixes that can change prior scores: test questions could leak into their own few-shot… https://benchmark-radar.org/blog/2026-08-31/ https://benchmark-radar.org/blog/2026-08-31/ Mon, 31 Aug 2026 00:00:00 GMT PCFBench reflects a recurring push in this captured feed to inspect an agent’s process rather than only its final answer. It separately tests… https://benchmark-radar.org/blog/2026-08-30/ https://benchmark-radar.org/blog/2026-08-30/ Sun, 30 Aug 2026 00:00:00 GMT The new INSIDER LLM Detection Benchmark evaluates models that may take harmful actions by comparing the model’s self-reported action log with an… https://benchmark-radar.org/blog/2026-08-29/ https://benchmark-radar.org/blog/2026-08-29/ Sat, 29 Aug 2026 00:00:00 GMT The new NBPO benchmark-generations dataset publishes every decoded model response used in three judge-based comparisons, allowing another evaluator to… https://benchmark-radar.org/blog/2026-08-28/ https://benchmark-radar.org/blog/2026-08-28/ Fri, 28 Aug 2026 00:00:00 GMT The new Same Model, Different Harness study holds the coding model and tasks fixed while changing how the agent harness manages conversation history and… https://benchmark-radar.org/blog/2026-08-27/ https://benchmark-radar.org/blog/2026-08-27/ Thu, 27 Aug 2026 00:00:00 GMT OpenCompass v0.5.4 is a substantive harness update, adding native VLMEvalKit-based multimodal evaluation, multi-round inference with the Multi-IF… https://benchmark-radar.org/blog/2026-08-26/ https://benchmark-radar.org/blog/2026-08-26/ Wed, 26 Aug 2026 00:00:00 GMT Across this captured feed, three new agent benchmarks make the evaluated unit an interactive model-plus-runtime system rather than a final answer… https://benchmark-radar.org/blog/2026-08-25/ https://benchmark-radar.org/blog/2026-08-25/ Tue, 25 Aug 2026 00:00:00 GMT New release SUSVIBES evaluates 12 coding-agent settings on 186 real-world feature requests for which human developers previously committed vulnerable… https://benchmark-radar.org/blog/2026-08-24/ https://benchmark-radar.org/blog/2026-08-24/ Mon, 24 Aug 2026 00:00:00 GMT No material GPT insight: No category moved far enough, persistently enough, or across enough independent sources to support a decision-useful finding in… https://benchmark-radar.org/blog/2026-08-23/ https://benchmark-radar.org/blog/2026-08-23/ Sun, 23 Aug 2026 00:00:00 GMT No material GPT insight: No category moved far enough, persistently enough, or across enough independent sources to support a material finding in today’s… https://benchmark-radar.org/blog/2026-08-22/ https://benchmark-radar.org/blog/2026-08-22/ Sat, 22 Aug 2026 00:00:00 GMT No material GPT insight: No material pattern was supported: the captured items did not show a sufficiently large, persistent, cross-source shift. Only 19… https://benchmark-radar.org/blog/2026-08-21/ https://benchmark-radar.org/blog/2026-08-21/ Fri, 21 Aug 2026 00:00:00 GMT No material GPT insight: No material pattern cleared the feed’s persistence and cross-source thresholds today. Only 65 of 198 corpus evidence records were… https://benchmark-radar.org/blog/2026-08-20/ https://benchmark-radar.org/blog/2026-08-20/ Thu, 20 Aug 2026 00:00:00 GMT No material GPT insight: No category changed far enough, persistently enough, and across enough sources to support a decision-useful pattern in this… https://benchmark-radar.org/blog/2026-08-19/ https://benchmark-radar.org/blog/2026-08-19/ Wed, 19 Aug 2026 00:00:00 GMT No material GPT insight: No category changed far enough, persistently enough, or across enough sources to support a material finding in this captured… https://benchmark-radar.org/blog/2026-08-18/ https://benchmark-radar.org/blog/2026-08-18/ Tue, 18 Aug 2026 00:00:00 GMT No material GPT insight: No category change was sufficiently large, persistent, and cross-source to support a material finding. Tracked metric movement… https://benchmark-radar.org/blog/2026-08-17/ https://benchmark-radar.org/blog/2026-08-17/ Mon, 17 Aug 2026 00:00:00 GMT Several new releases in the captured feed redesign evaluation around conditions hidden by static averages: evolving evidence and temporal cutoffs… https://benchmark-radar.org/blog/2026-08-16/ https://benchmark-radar.org/blog/2026-08-16/ Sun, 16 Aug 2026 00:00:00 GMT Agent benchmarking in the captured evidence is being designed around controlled execution, not just task sets: a new arena adds side-by-side and blind… https://benchmark-radar.org/blog/2026-08-15/ https://benchmark-radar.org/blog/2026-08-15/ Sat, 15 Aug 2026 00:00:00 GMT In today’s captured feed, three independent new releases evaluate agents as operational systems: LCAB preserves complete repair sessions and hardware… https://benchmark-radar.org/blog/2026-08-14/ https://benchmark-radar.org/blog/2026-08-14/ Fri, 14 Aug 2026 00:00:00 GMT Two new paper releases in this captured feed question whether aggregate agent scores measure deployable capability: one reports task interactions… https://benchmark-radar.org/blog/2026-08-13/ https://benchmark-radar.org/blog/2026-08-13/ Thu, 13 Aug 2026 00:00:00 GMT Across today’s captured new releases, several benchmarks require inspectable process evidence rather than scoring only final outcomes: replayable… https://benchmark-radar.org/blog/2026-08-12/ https://benchmark-radar.org/blog/2026-08-12/ Wed, 12 Aug 2026 00:00:00 GMT Several new releases in this captured feed bind benchmarks to deployment context: web-access APIs are compared across quality, latency, cost, and error… https://benchmark-radar.org/blog/2026-08-11/ https://benchmark-radar.org/blog/2026-08-11/ Tue, 11 Aug 2026 00:00:00 GMT Multiple new releases in the captured feed move agent evaluation beyond final-task success toward inspecting collaboration graphs, requirement recovery… https://benchmark-radar.org/blog/2026-08-10/ https://benchmark-radar.org/blog/2026-08-10/ Mon, 10 Aug 2026 00:00:00 GMT Several new releases in this captured feed evaluate whether a system reached an answer through valid evidence or execution, not merely whether the answer… https://benchmark-radar.org/blog/2026-08-08/ https://benchmark-radar.org/blog/2026-08-08/ Sat, 08 Aug 2026 00:00:00 GMT Across this feed’s new releases, benchmark design is moving toward specialized, structurally difficult tasks: open-ended scientific extraction… https://benchmark-radar.org/blog/2026-08-07/ https://benchmark-radar.org/blog/2026-08-07/ Fri, 07 Aug 2026 00:00:00 GMT Four newly released agent benchmarks in this captured feed expand evaluation beyond terminal task success: acquisition-stage privacy, learning across… https://benchmark-radar.org/blog/2026-08-06/ https://benchmark-radar.org/blog/2026-08-06/ Thu, 06 Aug 2026 00:00:00 GMT Across this captured feed, four newly released benchmarks evaluate behavior beyond a static, well-specified task: proactive bug discovery without issue… https://benchmark-radar.org/blog/2026-08-05/ https://benchmark-radar.org/blog/2026-08-05/ Wed, 05 Aug 2026 00:00:00 GMT Agentic artifacts rose to 26.3% of our captured feed over the last 5 days, against a 13.6% baseline across the prior 4 days (+12.7 percentage points). https://benchmark-radar.org/blog/2026-08-04/ https://benchmark-radar.org/blog/2026-08-04/ Tue, 04 Aug 2026 00:00:00 GMT Benchmark Radar collected 212 evidence observations from 3 sources on 2026-08-04. https://benchmark-radar.org/blog/2026-08-03/ https://benchmark-radar.org/blog/2026-08-03/ Mon, 03 Aug 2026 00:00:00 GMT Benchmark Radar collected 152 evidence observations from 3 sources on 2026-08-03. https://benchmark-radar.org/blog/2026-08-02/ https://benchmark-radar.org/blog/2026-08-02/ Sun, 02 Aug 2026 00:00:00 GMT Benchmark Radar collected 89 evidence observations from 2 sources on 2026-08-02. https://benchmark-radar.org/blog/2026-08-01/ https://benchmark-radar.org/blog/2026-08-01/ Sat, 01 Aug 2026 00:00:00 GMT Benchmark Radar collected 69 evidence observations from 2 sources on 2026-08-01. https://benchmark-radar.org/blog/2026-07-31/ https://benchmark-radar.org/blog/2026-07-31/ Fri, 31 Jul 2026 00:00:00 GMT Benchmark Radar collected 60 evidence observations from 2 sources on 2026-07-31. https://benchmark-radar.org/blog/2026-07-30/ https://benchmark-radar.org/blog/2026-07-30/ Thu, 30 Jul 2026 00:00:00 GMT Benchmark Radar collected 200 evidence observations from 3 sources on 2026-07-30. https://benchmark-radar.org/blog/2026-07-29/ https://benchmark-radar.org/blog/2026-07-29/ Wed, 29 Jul 2026 00:00:00 GMT Benchmark Radar collected 201 evidence observations from 3 sources on 2026-07-29. https://benchmark-radar.org/blog/2026-07-28/ https://benchmark-radar.org/blog/2026-07-28/ Tue, 28 Jul 2026 00:00:00 GMT Benchmark Radar collected 186 evidence observations from 3 sources on 2026-07-28. https://benchmark-radar.org/blog/2026-07-27/ https://benchmark-radar.org/blog/2026-07-27/ Mon, 27 Jul 2026 00:00:00 GMT Benchmark Radar collected 30 evidence observations from 2 sources on 2026-07-27. https://benchmark-radar.org/blog/2026-07-26/ https://benchmark-radar.org/blog/2026-07-26/ Sun, 26 Jul 2026 00:00:00 GMT Benchmark Radar collected 30 evidence observations from 1 sources on 2026-07-26. https://benchmark-radar.org/blog/2026-07-25/ https://benchmark-radar.org/blog/2026-07-25/ Sat, 25 Jul 2026 00:00:00 GMT Benchmark Radar collected 32 evidence observations from 1 sources on 2026-07-25. https://benchmark-radar.org/blog/2026-07-24/ https://benchmark-radar.org/blog/2026-07-24/ Fri, 24 Jul 2026 00:00:00 GMT Benchmark Radar collected 32 evidence observations from 1 sources on 2026-07-24. https://benchmark-radar.org/blog/2026-07-23/ https://benchmark-radar.org/blog/2026-07-23/ Thu, 23 Jul 2026 00:00:00 GMT Benchmark Radar collected 20 evidence observations from 1 sources on 2026-07-23.