šŸ¦ Twitter Post Details

Viewing enriched Twitter post

@YuvrajS9886

Training Qwen2.5-0.5B-Instruct on Reddit post summarization with GRPO on my 3x Mac Minis ā€” trying combination of quality rewards with length penalty! Completed all of the following combination rewards! >METEOR + BLEU >BLEU + ROUGE-L >METEOR + ROUGE-L All the code and wandb charts in the comments --- Training Qwen2.5-0.5B-Instruct on Reddit post summarization with GRPO on my 3x Mac Minis ā€” trying combination of quality rewards with length penalty! Completed all of the following combination rewards! >METEOR + BLEU >BLEU + ROUGE-L >METEOR + ROUGE-L All the code and wandb charts in the comments --- Setup: 3x Mac Minis in a cluster running MLX. One node drives training using GRPO, two push rollouts via vLLM. Trained two variants: ā†’ length penalty only (baseline) ā†’ length penalty + quality reward (BLEU, METEOR and/or ROUGE-L ) --- Eval: LLM-as-a-Judge (gpt-5) Used DeepEval to build a judge pipeline scoring each summary on 4 axes: ā†’ Faithfulness ā€” no hallucinations vs. source ā†’ Coverage ā€” key points captured ā†’ Conciseness ā€” shorter, no redundancy ā†’ Clarity ā€” readable on its own

View on Twitter

šŸ“Š Media Metadata

{
  "score": 0.42,
  "score_components": {
    "author": 0.09,
    "engagement": 0.0,
    "quality": 0.12,
    "source": 0.135,
    "nlp": 0.05,
    "recency": 0.025
  },
  "scored_at": "2026-04-18T00:19:03.858378",
  "import_source": "api_import",
  "source_tagged_at": "2026-04-18T00:19:03.858391",
  "enriched": true,
  "enriched_at": "2026-04-18T00:19:03.858393"
}

šŸ”§ Raw API Response

{
  "type": "tweet",
  "id": "2045137956848226677",
  "url": "https://x.com/YuvrajS9886/status/2045137956848226677",
  "twitterUrl": "https://twitter.com/YuvrajS9886/status/2045137956848226677",
  "text": "Training Qwen2.5-0.5B-Instruct on Reddit post summarization with GRPO on my 3x Mac Minis ā€” trying combination of quality rewards with length penalty!\n\nCompleted all of the following combination rewards!\n\n>METEOR + BLEU\n>BLEU + ROUGE-L\n>METEOR + ROUGE-L\n\nAll the code and wandb charts in the comments\n\n---\n\nTraining Qwen2.5-0.5B-Instruct on Reddit post summarization with GRPO on my 3x Mac Minis ā€” trying combination of quality rewards with length penalty!\n\nCompleted all of the following combination rewards!\n\n>METEOR + BLEU\n>BLEU + ROUGE-L\n>METEOR + ROUGE-L\n\nAll the code and wandb charts in the comments\n\n---\n\nSetup: 3x Mac Minis in a cluster running MLX.\n\nOne node drives training using GRPO, two push rollouts via vLLM. Trained two variants:\n\nā†’ length penalty only (baseline)\nā†’ length penalty + quality reward (BLEU, METEOR and/or ROUGE-L )\n\n---\n\nEval:\nLLM-as-a-Judge (gpt-5)\nUsed DeepEval to build a judge pipeline scoring each summary on 4 axes:\nā†’ Faithfulness ā€” no hallucinations vs. source\nā†’ Coverage ā€” key points captured\nā†’ Conciseness ā€” shorter, no redundancy\nā†’ Clarity ā€” readable on its own",
  "source": "Twitter for iPhone",
  "retweetCount": 0,
  "replyCount": 1,
  "likeCount": 2,
  "quoteCount": 1,
  "viewCount": 182,
  "createdAt": "Fri Apr 17 13:51:00 +0000 2026",
  "lang": "en",
  "bookmarkCount": 0,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2045137956848226677",
  "displayTextRange": [
    0,
    285
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "YuvrajS9886",
    "url": "https://x.com/YuvrajS9886",
    "twitterUrl": "https://twitter.com/YuvrajS9886",
    "id": "1756343897297764352",
    "name": "Yuvraj Singh",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1880213277814452225/D9ex676D_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/1756343897297764352/1769939620",
    "description": "",
    "location": "India ",
    "followers": 2921,
    "following": 761,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Sat Feb 10 15:46:50 +0000 2024",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 16475,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 392,
    "statusesCount": 5846,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2035276406167642335"
    ],
    "profile_bio": {
      "description": "Intern @askalphaxiv \n ā€¢ Building cluster in my dorm room\n ā€¢ Ex - @turboml, @puch_ai '25, @iiserkol24,@UofMaryland ā€˜24\n ā€¢ Building smolhub on the side!",
      "entities": {
        "description": {
          "user_mentions": [
            {
              "id_str": "",
              "indices": [
                7,
                19
              ],
              "name": "",
              "screen_name": "askalphaxiv"
            },
            {
              "id_str": "",
              "indices": [
                65,
                73
              ],
              "name": "",
              "screen_name": "turboml"
            },
            {
              "id_str": "",
              "indices": [
                75,
                83
              ],
              "name": "",
              "screen_name": "puch_ai"
            },
            {
              "id_str": "",
              "indices": [
                89,
                100
              ],
              "name": "",
              "screen_name": "iiserkol24"
            },
            {
              "id_str": "",
              "indices": [
                101,
                113
              ],
              "name": "",
              "screen_name": "UofMaryland"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "linktr.ee/yuvrajsingh9886",
              "expanded_url": "https://linktr.ee/yuvrajsingh9886",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/nbznXuM0Iv"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {},
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": {
    "id": "1509381007950204928",
    "name": "Machine Learning",
    "custom_banner_media_url": null
  },
  "article": null
}