alphaXiv ✓

@askalphaxiv · joined 09 Nov 2023

High fidelity research

54 214Followers

90Following

2 859Posts total

240.3KViews on collected posts

Son gönderiler

alphaXiv ✓@askalphaxiv4 Sep

“HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?” Most agent benchmarks fix the harness and only evaluate the model. So this paper explores whether LLMs can build and improve the harness itself, including the execution loop, tools, context, recovery, and https:/ 1.5K views · 17 likes · 4 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv4 Sep

“Language Models Can Control Their Own Attention” Long-context LLMs waste compute by scanning the entire KV cache at every decoding step, even when only a small region matters. So this paper introduces Declarative Attention, where the model explicitly switches between global, h 3.5K views · 70 likes · 13 reposts · 8 replies Open on X →

alphaXiv ✓@askalphaxiv3 Sep

“Trace as State: Reasoning Traces as Conditional States for Long-Context Transformers” Normally, task state discovered late can’t influence how earlier context was encoded. So this paper from https://t.co/iZq3IqAgC9 proposes to reuse the model’s reasoning trace as task state, h 2.9K views · 37 likes · 4 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv3 Sep

GPT-6 Astra is now state-of-the-art on Terminal-Bench-Science 0.1, surpassing the recently released Fable 5.1 It's not a coincidence that Terminal-Bench Science was the first benchmark OpenAI and Anthropic reported in their model release blogs - this is something the labs are h 7.9K views · 96 likes · 17 reposts · 4 replies Open on X →

alphaXiv ✓@askalphaxiv3 Sep

“Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement” Coding agents right now struggle to build software continuously over long horizons. This paper wraps existing coding agents in repeated planning, coding, and independent testing loops, ht 1.6K views · 28 likes · 2 reposts · 1 replies Open on X →

alphaXiv ✓@askalphaxiv3 Sep

“SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers” Looped Transformers usually gain effective depth by spending extra compute, making it unclear whether looping itself helps. SMELT loops the middle half of an MoE Transformer twice while matching FLOPs, https://t. 1.9K views · 26 likes · 5 reposts · 1 replies Open on X →

alphaXiv ✓@askalphaxiv2 Sep

"Losing my religion" A beautifully written blog by professor @togelius on reckoning with the fear that AI will one day make human thought unnecessary. His proposal, complementary intelligence, argues that keeping humans in the loop is not a constraint on progress. Rather, it is 4.1K views · 47 likes · 6 reposts · 3 replies Open on X →

alphaXiv ✓@askalphaxiv2 Sep

“Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall” Knowledge distillation helps reasoning during mid-training, but surprisingly hurts factual recall because teachers are much less confident on the remaining knowledge-heavy tokens. So this paper ht 3.4K views · 57 likes · 5 reposts · 1 replies Open on X →

alphaXiv ✓@askalphaxiv2 Sep

@askalphaxiv Double the score jump wow 256 views · 0 likes · 0 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv1 Sep

OpenResearch repo: https://t.co/TfgOphzm1u 2.5K views · 18 likes · 1 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv1 Sep

Fable 5.1 is now state-of-the-art for autoresearch Fable 5.1 scores 52.6% Terminal-Bench-Science 0.1, significantly higher than Fable 5 (24.7%) and GPT-5.6 Sol (22.4%) Try it out now on OpenResearch: https://t.co/n0w1EYq4t3 https://t.co/CxqAgieWAZ 38.1K views · 574 likes · 69 reposts · 10 replies Open on X →

alphaXiv ✓@askalphaxiv27 Aug

@askalphaxiv @tinkerapi Enjoyed collaborating on this one 🤝🤝 @rehaanahmad171 596 views · 7 likes · 0 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv27 Aug

Stay tuned to hear more about the research we've done on OpenResearch using Tinker 👀 OpenResearch repo: https://t.co/TfgOphzm1u 2.7K views · 27 likes · 4 reposts · 1 replies Open on X →

alphaXiv ✓@askalphaxiv27 Aug

Introducing autoresearch for arXiv papers: replicate and experiment on any arXiv paper using an army of Claude/Codex agents For post-training, your agents can use @tinkerapi to launch concurrent RL runs while you watch your experiment tree grow Try it out now on OpenResearch: h 60.5K views · 540 likes · 73 reposts · 19 replies Open on X →

alphaXiv ✓@askalphaxiv12 May

@askalphaxiv A 4B specialist matching Sonnet on a scoped task. That's the jar thesis. I've been building the pipeline that produces these automatically. Quality-gated distillation -- Pappy verifies every output, Moonshiner trains the specialist from the verified runs. The small m 1.4K views · 9 likes · 0 reposts · 3 replies Open on X →

alphaXiv ✓@askalphaxiv12 May

You can also check out our training scripts here: https://t.co/XVB4Nj1X5c Thanks to @sumanthrh and @charlie_ruan from the SkyRL team for their responsiveness in resolving issues and for providing an amazing RL library for the community! We'd also like to thank @a1zhang for his 3.9K views · 36 likes · 2 reposts · 0 replies Open on X →

alphaXiv ✓@askalphaxiv12 May

Checkout the blog post here: https://t.co/vOSPtXGQxx 4.8K views · 57 likes · 7 reposts · 1 replies Open on X →

alphaXiv ✓@askalphaxiv12 May

Reinforcing Recursive Language Models Can a 4B model learn to recursively call itself to answer hard long-context questions? We RL fine-tuned a small model to behave as a native RLM. On evidence selection across scientific papers, our 4B RLM matches Sonnet 4.6 in quality http 98.6K views · 597 likes · 84 reposts · 15 replies Open on X →

Aynı büyüklükteki hesaplara karşı

Son 90 güne ait 14 gönderi, 10K–100K takipçi aralığıyla yan yana. aynı büyüklükteki hesaplardan daha çok kişiye gösteriliyor.

Medyan görüntülenme2 777bu hesap92410K–100K için medyan

Erişim, %5.12%bu hesap3.62%10K–100K için medyan

Etkileşim, %1.43%bu hesap1.52%10K–100K için medyan

Ölçüt	Bu hesap	10K–100K için medyan	Oran
Gönderi başına medyan görüntülenme	2 777	924	3.01×
Erişim (görüntülenme ÷ takipçi)	5.12%	3.62%	1.41×
Etkileşim oranı	1.43%	1.52%	0.94×

Bu aralıktaki diğer hesaplar → Başka bir hesapla karşılaştır → Bu kıyas değerleri nasıl kuruluyor →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

60.5K27 Aug

2.7K

596

38.1K1 Sep

2.5K

2562 Sep

3.4K

4.1K

1.9K3 Sep

1.6K

7.9K

2.9K

3.5K4 Sep

1.5K

Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

1.06%27 Aug

1.20%

1.17%

1.73%1 Sep

0.81%

0.00%2 Sep

1.84%

1.40%

1.70%3 Sep

1.94%

1.49%

1.46%

2.61%4 Sep

1.40%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes52.8%2 243 in total

Reposts7.0%296 in total

Replies1.6%67 in total

Quotes0.8%32 in total

Bookmarks37.9%1 613 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Benzer hesaplar

Frequently asked questions about @askalphaxiv

How many followers does @askalphaxiv have?

54 214 followers.

How many posts has @askalphaxiv published?

2 859 posts.

How do posts by @askalphaxiv perform?

Collected posts: 18. Views: 240 304. Likes: 2 243.