---
title: "AlphaSignal Post Training: Dpo News"
url: "https://alphasignal.ai/topic/post-training/dpo"
description: "Latest AI and machine learning news about Dpo (Post Training) on AlphaSignal."
---

# AlphaSignal Post Training: Dpo News

Latest AI and machine learning news about Dpo (Post Training) on AlphaSignal.

## Articles

- [Cohere's North Small Translate Beats DeepL Across 50 Languages With Downloadable Weights](https://alphasignal.ai/news/cohere-s-north-small-translate-beats-deepl-across-50-languages-with.md) (2026-10-03)
- [Ai2's AstaBrief 8B Writes Cited Research Reports 3.5x Faster in One Pass](https://alphasignal.ai/news/ai2-s-astabrief-8b-writes-cited-research-reports-3-5x-faster-in-one-pass.md) (2026-10-02)
- [Arena's FLUX.2-dev Beats Every Open-Source Image Model With 66% Win Rate](https://alphasignal.ai/news/arena-s-flux-2-dev-beats-every-open-source-image-model-with-66-win-rate.md) (2026-10-02)
- [ETH Zurich's Adaptive Probes Beat DPO at Safety Without Breaking AI Transparency](https://alphasignal.ai/news/eth-zurich-s-adaptive-probes-beat-dpo-at-safety-without-breaking-ai-transparency.md) (2026-09-29)
- [Interactive TTS Teaches Voice AI to Whisper, Shout, and Grieve on Cue](https://alphasignal.ai/news/interactive-tts-teaches-voice-ai-to-whisper-shout-and-grieve-on-cue.md) (2026-09-22)
- [Goodfire's Silico Lets an Agent Crack Open AI Black Boxes Autonomously](https://alphasignal.ai/news/goodfire-s-silico-lets-an-agent-crack-open-ai-black-boxes-autonomously.md) (2026-08-04)
- [LiquidAI's Antidoom Slashes AI Reasoning Doom Loops by 95% in Hours](https://alphasignal.ai/news/liquidai-s-antidoom-slashes-ai-reasoning-doom-loops-by-95-in-hours.md) (2026-07-07)
- [Zyphra's Zamba2-1.2B Beats Gemma2-2B While Using 6x Less Memory](https://alphasignal.ai/news/zyphra-s-zamba2-1-2b-beats-gemma2-2b-while-using-6x-less-memory.md) (2026-07-07)
- [H2O.ai's LLM Studio Lets Anyone Fine-Tune Open-Source AI Without Writing Code](https://alphasignal.ai/news/h2o-ai-s-llm-studio-lets-anyone-fine-tune-open-source-ai-without-writing-code.md) (2026-06-30)
- [Intelligent Internet's Zenith Jumps GPT-5.5 From 5th to 1st on Hardest Coding Benchmark](https://alphasignal.ai/news/intelligent-internet-s-zenith-jumps-gpt-5-5-from-5th-to-1st-on-hardest-coding.md) (2026-06-29)
- [Hugging Face's TRL Hits v1.0 With 3M Monthly Downloads and a Stability Promise](https://alphasignal.ai/news/hugging-face-s-trl-hits-v1-0-with-3m-monthly-downloads-and-a-stability-promise.md) (2026-06-26)
- [Google's Tunix Unifies the Messy Open-Source Post-Training Stack in JAX](https://alphasignal.ai/news/google-s-tunix-unifies-the-messy-open-source-post-training-stack-in-jax.md) (2026-06-19)
- [AllenAI Opens the Full Training Stack Behind Tulu 3 to Beat GPT-4o-mini](https://alphasignal.ai/news/allenai-opens-the-full-training-stack-behind-tulu-3-to-beat-gpt-4o-mini.md) (2026-06-18)
- [Goodfire's Silico Catches Hidden AI Training Bugs Before a Single Step Runs](https://alphasignal.ai/news/goodfire-s-silico-catches-hidden-ai-training-bugs-before-a-single-step-runs.md) (2026-06-11)
- [Tencent's UniRL Open-Sources One RL Loop for Every AI Model Type](https://alphasignal.ai/news/tencent-s-unirl-open-sources-one-rl-loop-for-every-ai-model-type.md) (2026-06-09)
- [Loyola's TLA-Prover Beats Untuned AI at Formal Verification by 3.5x](https://alphasignal.ai/news/loyola-s-tla-prover-beats-untuned-ai-at-formal-verification-by-3-5x.md) (2026-06-04)
- [Amazon's LitMT-8B Beats a 235B Model at Literary Translation](https://alphasignal.ai/news/amazon-s-litmt-8b-beats-a-235b-model-at-literary-translation.md) (2026-06-04)
- [Iowa State Teaches AI to Reason Over Databases by Showing It Wrong Answers](https://alphasignal.ai/news/iowa-state-teaches-ai-to-reason-over-databases-by-showing-it-wrong-answers.md) (2026-06-03)
- [S-SPPO Fixes a Hidden Flaw That Was Quietly Breaking AI Alignment](https://alphasignal.ai/news/s-sppo-fixes-a-hidden-flaw-that-was-quietly-breaking-ai-alignment.md) (2026-06-01)
- [Deakin's SPaCe Matches Full-Dataset AI Training With Just 100 Examples](https://alphasignal.ai/news/deakin-s-space-matches-full-dataset-ai-training-with-just-100-examples.md) (2026-04-16)
- [Discrete Guidance Matching Fixes Broken AI Steering With 1.6x Faster Sampling](https://alphasignal.ai/news/discrete-guidance-matching-fixes-broken-ai-steering-with-1-6x-faster-sampling.md) (2026-04-15)
- [Meituan Trains AI Agents to Stop Repeating the Same Tool Mistakes](https://alphasignal.ai/news/meituan-trains-ai-agents-to-stop-repeating-the-same-tool-mistakes.md) (2026-04-15)
- [Haocheng Lu's Hard-Negative DPO Pushes Qwen2.5 Math Scores up 3.8 Points](https://alphasignal.ai/news/haocheng-lu-s-hard-negative-dpo-pushes-qwen2-5-math-scores-up-3-8-points.md) (2026-04-14)
- [Huawei's rDPO Beats Standard Vision AI Training With Per-Image Scoring Checklists](https://alphasignal.ai/news/huawei-s-rdpo-beats-standard-vision-ai-training-with-per-image-scoring.md) (2026-04-14)
- [STOMP Beats Multi-Objective AI Alignment Where Linear Mixing Always Fails](https://alphasignal.ai/news/stomp-beats-multi-objective-ai-alignment-where-linear-mixing-always-fails.md) (2026-04-14)
- [PsyCo-Llama3-8B Beats GPT-4o at Mental Health Counseling With 87% Win Rate](https://alphasignal.ai/news/psyco-llama3-8b-beats-gpt-4o-at-mental-health-counseling-with-87-win-rate.md) (2026-04-13)
- [CMU's GoodPoint Trains an 8B Model to Beat Gemini at Reviewing Research Papers](https://alphasignal.ai/news/cmu-s-goodpoint-trains-an-8b-model-to-beat-gemini-at-reviewing-research-papers.md) (2026-04-13)
- [Technion and Nvidia Catch LLMs Lying About Their Own Reasoning](https://alphasignal.ai/news/technion-and-nvidia-catch-llms-lying-about-their-own-reasoning.md) (2026-04-12)
- [USC's DreamPlan Teaches Robots to Handle Cloth and Ropes, Nearly Doubling AI Performance](https://alphasignal.ai/news/usc-s-dreamplan-teaches-robots-to-handle-cloth-and-ropes-nearly-doubling-ai.md) (2026-03-17)
- [Peking University's EvolvR Beats GPT-o3 at Judging AI-Written Stories](https://alphasignal.ai/news/peking-university-s-evolvr-beats-gpt-o3-at-judging-ai-written-stories.md) (2026-03-14)
- [Apple's AdaBoN Cuts AI Inference Waste by 20% Without Retraining](https://alphasignal.ai/news/apple-s-adabon-cuts-ai-inference-waste-by-20-without-retraining.md) (2026-03-13)
- [ByteDance's LibraGen Beats Kling and Vidu With Just 9,000 Training Samples](https://alphasignal.ai/news/bytedance-s-libragen-beats-kling-and-vidu-with-just-9-000-training-samples.md) (2026-03-13)
- [New Theory Formally Proves When AI Models Forget Old Skills During Training](https://alphasignal.ai/news/new-theory-formally-proves-when-ai-models-forget-old-skills-during-training.md) (2026-03-12)
- [Imperial College's AbSteering Turns Video AI Into a Radiology Expert](https://alphasignal.ai/news/imperial-college-s-absteering-turns-video-ai-into-a-radiology-expert.md) (2026-03-12)
- [Harbin Institute's JudgeBiasBench Exposes 12 Ways AI Judges Corrupt RLHF Training](https://alphasignal.ai/news/harbin-institute-s-judgebiasbench-exposes-12-ways-ai-judges-corrupt-rlhf.md) (2026-03-09)
- [Meta and CMU's BTPO Fixes the Broken Math Behind AI Preference Training](https://alphasignal.ai/news/meta-and-cmu-s-btpo-fixes-the-broken-math-behind-ai-preference-training.md) (2026-03-09)
- [Researchers Prove AI Leaderboards Are Broken and Tune-Before-Test Fixes Them](https://alphasignal.ai/news/researchers-prove-ai-leaderboards-are-broken-and-tune-before-test-fixes-them.md) (2026-03-09)
- [ANU Beats ReinMax With Two Gradient Estimators That Train Discrete AI Models Better](https://alphasignal.ai/news/anu-beats-reinmax-with-two-gradient-estimators-that-train-discrete-ai-models.md) (2026-03-09)
- [Peking University's CDRRM Fixes AI Reward Hacking With Evidence-Grounded Rubrics](https://alphasignal.ai/news/peking-university-s-cdrrm-fixes-ai-reward-hacking-with-evidence-grounded-rubrics.md) (2026-03-09)
- [ByteDance's SamS Makes AI Alignment Training 12% Smarter by Picking Better Examples](https://alphasignal.ai/news/bytedance-s-sams-makes-ai-alignment-training-12-smarter-by-picking-better.md) (2026-03-09)
- [Huawei's EDU-PRM Matches Top AI Reasoning Checks With 1.5% of the Data](https://alphasignal.ai/news/huawei-s-edu-prm-matches-top-ai-reasoning-checks-with-1-5-of-the-data.md) (2026-03-09)
- [TinyLlama-1.1B Beats Bigger Models Running on Just 637MB](https://alphasignal.ai/news/tinyllama-1-1b-beats-bigger-models-running-on-just-637mb.md) (2023-12-30)
- [Hugging Face's Zephyr-7B Beats Llama 2 70B With a Smarter Training Recipe](https://alphasignal.ai/news/hugging-face-s-zephyr-7b-beats-llama-2-70b-with-a-smarter-training-recipe.md) (2023-10-26)