Hackernews posts about HumanEval

RSS feed for HumanEval

01
newShow HN: I trained a 160M code model on one GPU and got 22.56% on HumanEvaljakubcwirlej.com
▲ 1jakcwi 7d discuss
02
newMaincoder-1B – an open 1B-parameter coding model with 76% HumanEvalhuggingface.co
▲ 4MainNews 311d 💬 3
03
newShow HN: S0 Tuning – +23.6pp on HumanEval by tuning state, not weightsgithub.com
▲ 2jacknotold 211d 💬 2
04
newLaunch HN: Strata (YC P25) – One MCP server for AI to handle thousands of tools
▲ 133wirehack 402d 💬 66
05
newLLM Benchmark: Frontier models now statistically indistinguishable
▲ 7js4ever 314d 💬 4
06
newShow HN: I built the LLM Comparison Tool I wish existedllm-stats.com
▲ 7JonathanChavez 665d 💬 3
07
newShow HN: Benchmark AI on your actual code (GPT-5, Claude, Grok, Gemini, o3)codelens.ai
▲ 7codelensai 384d discuss
08
newShow HN: European Swallow AI – Sonnet-quality coding at $2.60/M tokenswww.europeanswallowai.com
▲ 7joaquim_d 386d discuss
09
newDiffusion LLM may make most of the AI engineering stack obsolete
▲ 3victorpiles99 233d discuss
10
newShow HN: Loki Mode hit 99.67% SWE-Bench – MAF built a SaaS overnightgithub.com
▲ 2slogansand 296d 💬 5
11
newQ Evaluation Harness: open-source evals for LLMs on q/kdb+github.com
▲ 2erfan_mhi 441d discuss
12
newShow HN: AIBenchy – Independent AI Leaderboardaibenchy.com
▲ 1XCSme 255d 💬 1
13
newARCHE3-7B – Sparse Moe with SmartRouter and Foundation Curriculum Training
▲ 1OpenSynapseLabs 210d discuss
14
newShow HN: Kore – Stack based language where compiler is the reward functiongithub.com
▲ 1processorx 263d discuss
15
newShow HN: Atlas: Independent Evals and Benchmarking for Generative AI Modelsapp.layerlens.ai
▲ 1Arch223 534d discuss
16
newGPT5 is the best coding LLM because other LLMs admit it?
▲ 1adinhitlore 430d 💬 8